/[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 85 by nigel, Sat Feb 24 21:41:13 2007 UTC revision 276 by ph10, Wed Nov 21 15:39:20 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    
# Line 53  used by pcretest. DEBUG is not defined w Line 61  used by pcretest. DEBUG is not defined w
61  #endif  #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
79    
80  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
81  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
82  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
83  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
84  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
85    so this number is very generous.
86    
87    The same workspace is used during the second, actual compile phase for
88    remembering forward references to groups so that they can be filled in at the
89    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
90    is 4 there is plenty of room. */
91    
92  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
93    
94    
95  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 72  are simple data values; negative values Line 97  are simple data values; negative values
97  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
98  is invalid. */  is invalid. */
99    
100  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */
101  static const short int escapes[] = {  static const short int escapes[] = {
102       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */
103       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */
104     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */
105       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */
106  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */
107  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */
108     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */
109       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */
110  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */
111       0,      0, -ESC_z                                            /* x - z */       0,      0, -ESC_z                                            /* x - z */
112  };  };
113    
114  #else         /* This is the "abnormal" table for EBCDIC systems */  #else           /* This is the "abnormal" table for EBCDIC systems */
115  static const short int escapes[] = {  static const short int escapes[] = {
116  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
117  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 96  static const short int escapes[] = { Line 121  static const short int escapes[] = {
121  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
122  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
123  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
124  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
125  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
126  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
127  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
128  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
129  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
130  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
131  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
132  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
133  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
134  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
135  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
136  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
137  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
138  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 115  static const short int escapes[] = { Line 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
144  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
145  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. */
146    
147  static const char *const posix_names[] = {  typedef struct verbitem {
148    "alpha", "lower", "upper",    int   len;
149    "alnum", "ascii", "blank", "cntrl", "digit", "graph",    int   op;
150    "print", "punct", "space", "word",  "xdigit" };  } verbitem;
151    
152    static const char verbnames[] =
153      "ACCEPT\0"
154      "COMMIT\0"
155      "F\0"
156      "FAIL\0"
157      "PRUNE\0"
158      "SKIP\0"
159      "THEN";
160    
161    static verbitem verbs[] = {
162      { 6, OP_ACCEPT },
163      { 6, OP_COMMIT },
164      { 1, OP_FAIL },
165      { 4, OP_FAIL },
166      { 5, OP_PRUNE },
167      { 4, OP_SKIP  },
168      { 4, OP_THEN  }
169    };
170    
171    static int verbcount = sizeof(verbs)/sizeof(verbitem);
172    
173    
174    /* Tables of names of POSIX character classes and their lengths. The names are
175    now all in a single string, to reduce the number of relocations when a shared
176    library is dynamically loaded. The list of lengths is terminated by a zero
177    length entry. The first three must be alpha, lower, upper, as this is assumed
178    for handling case independence. */
179    
180    static const char posix_names[] =
181      "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"
182      "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"
183      "word\0"   "xdigit";
184    
185  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
186    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 };
187    
188  /* 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
189  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
190  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
191    characters are removed, and for [:alpha:] and [:alnum:] the underscore
192    character is removed. The triples in the table consist of the base map offset,
193    second map offset or -1 if no second map, and a non-negative value for map
194    addition or a negative value for map subtraction (if there are two maps). The
195    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
196    remove vertical space characters, 2 => remove underscore. */
197    
198  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
199    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
200    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
201    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
202    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
203    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
204    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
205    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
206    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
207    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
208    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
209    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
210    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
211    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
212    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
213  };  };
214    
215    
216  /* The texts of compile-time error messages. These are "char *" because they  #define STRING(a)  # a
217  are passed to the outside world. */  #define XSTRING(s) STRING(s)
218    
219  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
220    "no error",  are passed to the outside world. Do not ever re-use any error number, because
221    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
222    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
223    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
224    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
225    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
226    simply count through to the one we want - this isn't a performance issue
227    because these strings are used only when there is a compilation error. */
228    
229    static const char error_texts[] =
230      "no error\0"
231      "\\ at end of pattern\0"
232      "\\c at end of pattern\0"
233      "unrecognized character follows \\\0"
234      "numbers out of order in {} quantifier\0"
235    /* 5 */    /* 5 */
236    "number too big in {} quantifier",    "number too big in {} quantifier\0"
237    "missing terminating ] for character class",    "missing terminating ] for character class\0"
238    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
239    "range out of order in character class",    "range out of order in character class\0"
240    "nothing to repeat",    "nothing to repeat\0"
241    /* 10 */    /* 10 */
242    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
243    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
244    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
245    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
246    "missing )",    "missing )\0"
247    /* 15 */    /* 15 */
248    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
249    "erroffset passed as NULL",    "erroffset passed as NULL\0"
250    "unknown option bit(s) set",    "unknown option bit(s) set\0"
251    "missing ) after comment",    "missing ) after comment\0"
252    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
253    /* 20 */    /* 20 */
254    "regular expression too large",    "regular expression is too large\0"
255    "failed to get memory",    "failed to get memory\0"
256    "unmatched parentheses",    "unmatched parentheses\0"
257    "internal error: code overflow",    "internal error: code overflow\0"
258    "unrecognized character after (?<",    "unrecognized character after (?<\0"
259    /* 25 */    /* 25 */
260    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
261    "malformed number after (?(",    "malformed number or name after (?(\0"
262    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
263    "assertion expected after (?(",    "assertion expected after (?(\0"
264    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
265    /* 30 */    /* 30 */
266    "unknown POSIX class name",    "unknown POSIX class name\0"
267    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
268    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
269    "spare error",    "spare error\0"  /** DEAD **/
270    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
271    /* 35 */    /* 35 */
272    "invalid condition (?(0)",    "invalid condition (?(0)\0"
273    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
274    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
275    "number after (?C is > 255",    "number after (?C is > 255\0"
276    "closing ) for (?C expected",    "closing ) for (?C expected\0"
277    /* 40 */    /* 40 */
278    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
279    "unrecognized character after (?P",    "unrecognized character after (?P\0"
280    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
281    "two named groups have the same name",    "two named subpatterns have the same name\0"
282    "invalid UTF-8 string",    "invalid UTF-8 string\0"
283    /* 45 */    /* 45 */
284    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
285    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
286    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
287  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
288      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
289      /* 50 */
290      "repeated subpattern is too long\0"    /** DEAD **/
291      "octal value is greater than \\377 (not in UTF-8 mode)\0"
292      "internal error: overran compiling workspace\0"
293      "internal error: previously-checked referenced subpattern not found\0"
294      "DEFINE group contains more than one branch\0"
295      /* 55 */
296      "repeating a DEFINE group is not allowed\0"
297      "inconsistent NEWLINE options\0"
298      "\\g is not followed by a braced name or an optionally braced non-zero number\0"
299      "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"
300      "(*VERB) with an argument is not supported\0"
301      /* 60 */
302      "(*VERB) not recognized\0"
303      "number is too big\0"
304      "subpattern name expected\0"
305      "digit expected after (?+";
306    
307    
308  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 229  For convenience, we use the same bit def Line 321  For convenience, we use the same bit def
321    
322  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
323    
324  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */
325  static const unsigned char digitab[] =  static const unsigned char digitab[] =
326    {    {
327    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 265  static const unsigned char digitab[] = Line 357  static const unsigned char digitab[] =
357    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
358    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
359    
360  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else           /* This is the "abnormal" case, for EBCDIC systems */
361  static const unsigned char digitab[] =  static const unsigned char digitab[] =
362    {    {
363    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 279  static const unsigned char digitab[] = Line 371  static const unsigned char digitab[] =
371    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
372    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
373    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
374    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
375    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
376    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
377    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 313  static const unsigned char ebcdic_charta Line 405  static const unsigned char ebcdic_charta
405    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
406    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
407    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
408    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
409    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
410    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
411    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 340  static const unsigned char ebcdic_charta Line 432  static const unsigned char ebcdic_charta
432  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
433    
434  static BOOL  static BOOL
435    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
436      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
437    
438    
439    
440    /*************************************************
441    *            Find an error text                  *
442    *************************************************/
443    
444    /* The error texts are now all in one long string, to save on relocations. As
445    some of the text is of unknown length, we can't use a table of offsets.
446    Instead, just count through the strings. This is not a performance issue
447    because it happens only when there has been a compilation error.
448    
449    Argument:   the error number
450    Returns:    pointer to the error string
451    */
452    
453    static const char *
454    find_error_text(int n)
455    {
456    const char *s = error_texts;
457    for (; n > 0; n--) while (*s++ != 0);
458    return s;
459    }
460    
461    
462  /*************************************************  /*************************************************
# Line 351  static BOOL Line 465  static BOOL
465    
466  /* 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
467  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
468  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
469  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
470  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,
471    ptr is pointing at the \. On exit, it is on the final character of the escape
472    sequence.
473    
474  Arguments:  Arguments:
475    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 364  Arguments: Line 480  Arguments:
480    
481  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
482                   negative => a special escape sequence                   negative => a special escape sequence
483                   on error, errorptr is set                   on error, errorcodeptr is set
484  */  */
485    
486  static int  static int
487  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
488    int options, BOOL isclass)    int options, BOOL isclass)
489  {  {
490  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
491    const uschar *ptr = *ptrptr + 1;
492  int c, i;  int c, i;
493    
494    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
495    ptr--;                            /* Set pointer back to the last byte */
496    
497  /* 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. */
498    
 c = *(++ptr);  
499  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
500    
501  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
502  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
503  Otherwise further processing may be required. */  Otherwise further processing may be required. */
504    
505  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
506  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */
507  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
508    
509  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
510  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
511  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
512  #endif  #endif
513    
# Line 397  else if ((i = escapes[c - 0x48]) != 0) Line 516  else if ((i = escapes[c - 0x48]) != 0)
516  else  else
517    {    {
518    const uschar *oldptr;    const uschar *oldptr;
519      BOOL braced, negated;
520    
521    switch (c)    switch (c)
522      {      {
523      /* 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 410  else Line 531  else
531      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
532      break;      break;
533    
534        /* \g must be followed by a number, either plain or braced. If positive, it
535        is an absolute backreference. If negative, it is a relative backreference.
536        This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a
537        reference to a named group. This is part of Perl's movement towards a
538        unified syntax for back references. As this is synonymous with \k{name}, we
539        fudge it up by pretending it really was \k. */
540    
541        case 'g':
542        if (ptr[1] == '{')
543          {
544          const uschar *p;
545          for (p = ptr+2; *p != 0 && *p != '}'; p++)
546            if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;
547          if (*p != 0 && *p != '}')
548            {
549            c = -ESC_k;
550            break;
551            }
552          braced = TRUE;
553          ptr++;
554          }
555        else braced = FALSE;
556    
557        if (ptr[1] == '-')
558          {
559          negated = TRUE;
560          ptr++;
561          }
562        else negated = FALSE;
563    
564        c = 0;
565        while ((digitab[ptr[1]] & ctype_digit) != 0)
566          c = c * 10 + *(++ptr) - '0';
567    
568        if (c < 0)
569          {
570          *errorcodeptr = ERR61;
571          break;
572          }
573    
574        if (c == 0 || (braced && *(++ptr) != '}'))
575          {
576          *errorcodeptr = ERR57;
577          break;
578          }
579    
580        if (negated)
581          {
582          if (c > bracount)
583            {
584            *errorcodeptr = ERR15;
585            break;
586            }
587          c = bracount - (c - 1);
588          }
589    
590        c = -(ESC_REF + c);
591        break;
592    
593      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
594      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
595      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 431  else Line 611  else
611        c -= '0';        c -= '0';
612        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
613          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
614          if (c < 0)
615            {
616            *errorcodeptr = ERR61;
617            break;
618            }
619        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
620          {          {
621          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 451  else Line 636  else
636        }        }
637    
638      /* \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
639      larger first octal digit. */      larger first octal digit. The original code used just to take the least
640        significant 8 bits of octal numbers (I think this is what early Perls used
641        to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
642        than 3 octal digits. */
643    
644      case '0':      case '0':
645      c -= '0';      c -= '0';
646      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')
647          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - '0';
648      c &= 255;     /* Take least significant 8 bits */      if (!utf8 && c > 255) *errorcodeptr = ERR51;
649      break;      break;
650    
651      /* \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
652      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
653        treated as a data character. */
654    
655      case 'x':      case 'x':
656  #ifdef SUPPORT_UTF8      if (ptr[1] == '{')
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
657        {        {
658        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
659        register int count = 0;        int count = 0;
660    
661        c = 0;        c = 0;
662        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
663          {          {
664          int cc = *pt++;          register int cc = *pt++;
665            if (c == 0 && cc == '0') continue;     /* Leading zeroes */
666          count++;          count++;
667  #if !EBCDIC    /* ASCII coding */  
668    #ifndef EBCDIC  /* ASCII coding */
669          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= 'a') cc -= 32;               /* Convert to upper case */
670          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));
671  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
672          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */
673          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));
674  #endif  #endif
675          }          }
676    
677        if (*pt == '}')        if (*pt == '}')
678          {          {
679          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
680          ptr = pt;          ptr = pt;
681          break;          break;
682          }          }
683    
684        /* 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
685        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
686        }        }
 #endif  
687    
688      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
689    
690      c = 0;      c = 0;
691      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
692        {        {
693        int cc;                               /* Some compilers don't like ++ */        int cc;                               /* Some compilers don't like ++ */
694        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                        /* in initializers */
695  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
696        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= 'a') cc -= 32;              /* Convert to upper case */
697        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));
698  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
699        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= 'z') cc += 64;              /* Convert to upper case */
700        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));
701  #endif  #endif
702        }        }
703      break;      break;
704    
705      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
706        This coding is ASCII-specific, but then the whole concept of \cx is
707        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
708    
709      case 'c':      case 'c':
710      c = *(++ptr);      c = *(++ptr);
711      if (c == 0)      if (c == 0)
712        {        {
713        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
714        return 0;        break;
715        }        }
716    
717      /* 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 */  
718      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= 'a' && c <= 'z') c -= 32;
719      c ^= 0x40;      c ^= 0x40;
720  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
721      if (c >= 'a' && c <= 'z') c += 64;      if (c >= 'a' && c <= 'z') c += 64;
722      c ^= 0xC0;      c ^= 0xC0;
723  #endif  #endif
724      break;      break;
725    
726      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
727      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
728      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
729      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
730      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
731    
732      default:      default:
733      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 569  escape sequence. Line 759  escape sequence.
759  Argument:  Argument:
760    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
761    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
762      dptr           points to an int that is set to the detailed property value
763    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
764    
765  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
766  */  */
767    
768  static int  static int
769  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
770  {  {
771  int c, i, bot, top;  int c, i, bot, top;
772  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
773  char name[4];  char name[32];
774    
775  c = *(++ptr);  c = *(++ptr);
776  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
777    
778  *negptr = FALSE;  *negptr = FALSE;
779    
780  /* \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
781  preceded by ^ for negation. */  negation. */
782    
783  if (c == '{')  if (c == '{')
784    {    {
# Line 596  if (c == '{') Line 787  if (c == '{')
787      *negptr = TRUE;      *negptr = TRUE;
788      ptr++;      ptr++;
789      }      }
790    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
791      {      {
792      c = *(++ptr);      c = *(++ptr);
793      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
794      if (c == '}') break;      if (c == '}') break;
795      name[i] = c;      name[i] = c;
796      }      }
797    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;  
     }  
798    name[i] = 0;    name[i] = 0;
799    }    }
800    
# Line 628  top = _pcre_utt_size; Line 815  top = _pcre_utt_size;
815    
816  while (bot < top)  while (bot < top)
817    {    {
818    i = (bot + top)/2;    i = (bot + top) >> 1;
819    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
820    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
821        {
822        *dptr = _pcre_utt[i].value;
823        return _pcre_utt[i].type;
824        }
825    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
826    }    }
827    
 UNKNOWN_RETURN:  
828  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
829  *ptrptr = ptr;  *ptrptr = ptr;
830  return -1;  return -1;
# Line 750  return p; Line 940  return p;
940    
941    
942  /*************************************************  /*************************************************
943    *       Find forward referenced subpattern       *
944    *************************************************/
945    
946    /* This function scans along a pattern's text looking for capturing
947    subpatterns, and counting them. If it finds a named pattern that matches the
948    name it is given, it returns its number. Alternatively, if the name is NULL, it
949    returns when it reaches a given numbered subpattern. This is used for forward
950    references to subpatterns. We know that if (?P< is encountered, the name will
951    be terminated by '>' because that is checked in the first pass.
952    
953    Arguments:
954      ptr          current position in the pattern
955      count        current count of capturing parens so far encountered
956      name         name to seek, or NULL if seeking a numbered subpattern
957      lorn         name length, or subpattern number if name is NULL
958      xmode        TRUE if we are in /x mode
959    
960    Returns:       the number of the named subpattern, or -1 if not found
961    */
962    
963    static int
964    find_parens(const uschar *ptr, int count, const uschar *name, int lorn,
965      BOOL xmode)
966    {
967    const uschar *thisname;
968    
969    for (; *ptr != 0; ptr++)
970      {
971      int term;
972    
973      /* Skip over backslashed characters and also entire \Q...\E */
974    
975      if (*ptr == '\\')
976        {
977        if (*(++ptr) == 0) return -1;
978        if (*ptr == 'Q') for (;;)
979          {
980          while (*(++ptr) != 0 && *ptr != '\\');
981          if (*ptr == 0) return -1;
982          if (*(++ptr) == 'E') break;
983          }
984        continue;
985        }
986    
987      /* Skip over character classes */
988    
989      if (*ptr == '[')
990        {
991        while (*(++ptr) != ']')
992          {
993          if (*ptr == 0) return -1;
994          if (*ptr == '\\')
995            {
996            if (*(++ptr) == 0) return -1;
997            if (*ptr == 'Q') for (;;)
998              {
999              while (*(++ptr) != 0 && *ptr != '\\');
1000              if (*ptr == 0) return -1;
1001              if (*(++ptr) == 'E') break;
1002              }
1003            continue;
1004            }
1005          }
1006        continue;
1007        }
1008    
1009      /* Skip comments in /x mode */
1010    
1011      if (xmode && *ptr == '#')
1012        {
1013        while (*(++ptr) != 0 && *ptr != '\n');
1014        if (*ptr == 0) return -1;
1015        continue;
1016        }
1017    
1018      /* An opening parens must now be a real metacharacter */
1019    
1020      if (*ptr != '(') continue;
1021      if (ptr[1] != '?' && ptr[1] != '*')
1022        {
1023        count++;
1024        if (name == NULL && count == lorn) return count;
1025        continue;
1026        }
1027    
1028      ptr += 2;
1029      if (*ptr == 'P') ptr++;                      /* Allow optional P */
1030    
1031      /* We have to disambiguate (?<! and (?<= from (?<name> */
1032    
1033      if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&
1034           *ptr != '\'')
1035        continue;
1036    
1037      count++;
1038    
1039      if (name == NULL && count == lorn) return count;
1040      term = *ptr++;
1041      if (term == '<') term = '>';
1042      thisname = ptr;
1043      while (*ptr != term) ptr++;
1044      if (name != NULL && lorn == ptr - thisname &&
1045          strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1046        return count;
1047      }
1048    
1049    return -1;
1050    }
1051    
1052    
1053    
1054    /*************************************************
1055  *      Find first significant op code            *  *      Find first significant op code            *
1056  *************************************************/  *************************************************/
1057    
# Line 798  for (;;) Line 1100  for (;;)
1100    
1101      case OP_CALLOUT:      case OP_CALLOUT:
1102      case OP_CREF:      case OP_CREF:
1103      case OP_BRANUMBER:      case OP_RREF:
1104        case OP_DEF:
1105      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1106      break;      break;
1107    
# Line 843  for (;;) Line 1146  for (;;)
1146    {    {
1147    int d;    int d;
1148    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1149    switch (op)    switch (op)
1150      {      {
1151        case OP_CBRA:
1152      case OP_BRA:      case OP_BRA:
1153      case OP_ONCE:      case OP_ONCE:
1154      case OP_COND:      case OP_COND:
1155      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1156      if (d < 0) return d;      if (d < 0) return d;
1157      branchlength += d;      branchlength += d;
1158      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 885  for (;;) Line 1187  for (;;)
1187      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1188    
1189      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1190      case OP_CREF:      case OP_CREF:
1191        case OP_RREF:
1192        case OP_DEF:
1193      case OP_OPT:      case OP_OPT:
1194      case OP_CALLOUT:      case OP_CALLOUT:
1195      case OP_SOD:      case OP_SOD:
# Line 904  for (;;) Line 1207  for (;;)
1207    
1208      case OP_CHAR:      case OP_CHAR:
1209      case OP_CHARNC:      case OP_CHARNC:
1210        case OP_NOT:
1211      branchlength++;      branchlength++;
1212      cc += 2;      cc += 2;
1213  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 930  for (;;) Line 1234  for (;;)
1234    
1235      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1236      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1237        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1238      cc += 4;      cc += 4;
1239      break;      break;
1240    
# Line 937  for (;;) Line 1242  for (;;)
1242    
1243      case OP_PROP:      case OP_PROP:
1244      case OP_NOTPROP:      case OP_NOTPROP:
1245      cc++;      cc += 2;
1246      /* Fall through */      /* Fall through */
1247    
1248      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 1018  Returns:      pointer to the opcode for Line 1323  Returns:      pointer to the opcode for
1323  static const uschar *  static const uschar *
1324  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1325  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1326  for (;;)  for (;;)
1327    {    {
1328    register int c = *code;    register int c = *code;
1329    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1330    else if (c > OP_BRA)  
1331      /* XCLASS is used for classes that cannot be represented just by a bit
1332      map. This includes negated single high-valued characters. The length in
1333      the table is zero; the actual length is stored in the compiled code. */
1334    
1335      if (c == OP_XCLASS) code += GET(code, 1);
1336    
1337      /* Handle capturing bracket */
1338    
1339      else if (c == OP_CBRA)
1340      {      {
1341      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1342      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1343      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1344      }      }
1345    
1346      /* Otherwise, we can get the item's length from the table, except that for
1347      repeated character types, we have to test for \p and \P, which have an extra
1348      two bytes of parameters. */
1349    
1350    else    else
1351      {      {
1352      code += _pcre_OP_lengths[c];      switch(c)
1353          {
1354          case OP_TYPESTAR:
1355          case OP_TYPEMINSTAR:
1356          case OP_TYPEPLUS:
1357          case OP_TYPEMINPLUS:
1358          case OP_TYPEQUERY:
1359          case OP_TYPEMINQUERY:
1360          case OP_TYPEPOSSTAR:
1361          case OP_TYPEPOSPLUS:
1362          case OP_TYPEPOSQUERY:
1363          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1364          break;
1365    
1366  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1367          case OP_TYPEMINUPTO:
1368          case OP_TYPEEXACT:
1369          case OP_TYPEPOSUPTO:
1370          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1371          break;
1372          }
1373    
1374      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* Add in the fixed length from the table */
1375      by a multi-byte character. The length in the table is a minimum, so we have  
1376      to scan along to skip the extra bytes. All opcodes are less than 128, so we      code += _pcre_OP_lengths[c];
1377      can use relatively efficient code. */  
1378      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1379      a multi-byte character. The length in the table is a minimum, so we have to
1380      arrange to skip the extra bytes. */
1381    
1382    #ifdef SUPPORT_UTF8
1383      if (utf8) switch(c)      if (utf8) switch(c)
1384        {        {
1385        case OP_CHAR:        case OP_CHAR:
# Line 1051  for (;;) Line 1387  for (;;)
1387        case OP_EXACT:        case OP_EXACT:
1388        case OP_UPTO:        case OP_UPTO:
1389        case OP_MINUPTO:        case OP_MINUPTO:
1390          case OP_POSUPTO:
1391        case OP_STAR:        case OP_STAR:
1392        case OP_MINSTAR:        case OP_MINSTAR:
1393          case OP_POSSTAR:
1394        case OP_PLUS:        case OP_PLUS:
1395        case OP_MINPLUS:        case OP_MINPLUS:
1396          case OP_POSPLUS:
1397        case OP_QUERY:        case OP_QUERY:
1398        case OP_MINQUERY:        case OP_MINQUERY:
1399        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1400        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;  
1401        break;        break;
1402        }        }
1403  #endif  #endif
# Line 1092  Returns:      pointer to the opcode for Line 1424  Returns:      pointer to the opcode for
1424  static const uschar *  static const uschar *
1425  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1426  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1427  for (;;)  for (;;)
1428    {    {
1429    register int c = *code;    register int c = *code;
1430    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1431    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1432    else if (c > OP_BRA)  
1433      {    /* XCLASS is used for classes that cannot be represented just by a bit
1434      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1435      }    the table is zero; the actual length is stored in the compiled code. */
1436    
1437      if (c == OP_XCLASS) code += GET(code, 1);
1438    
1439      /* Otherwise, we can get the item's length from the table, except that for
1440      repeated character types, we have to test for \p and \P, which have an extra
1441      two bytes of parameters. */
1442    
1443    else    else
1444      {      {
1445      code += _pcre_OP_lengths[c];      switch(c)
1446          {
1447          case OP_TYPESTAR:
1448          case OP_TYPEMINSTAR:
1449          case OP_TYPEPLUS:
1450          case OP_TYPEMINPLUS:
1451          case OP_TYPEQUERY:
1452          case OP_TYPEMINQUERY:
1453          case OP_TYPEPOSSTAR:
1454          case OP_TYPEPOSPLUS:
1455          case OP_TYPEPOSQUERY:
1456          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1457          break;
1458    
1459  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1460          case OP_TYPEUPTO:
1461          case OP_TYPEMINUPTO:
1462          case OP_TYPEEXACT:
1463          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1464          break;
1465          }
1466    
1467        /* Add in the fixed length from the table */
1468    
1469        code += _pcre_OP_lengths[c];
1470    
1471      /* 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
1472      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
1473      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. */  
1474    
1475    #ifdef SUPPORT_UTF8
1476      if (utf8) switch(c)      if (utf8) switch(c)
1477        {        {
1478        case OP_CHAR:        case OP_CHAR:
# Line 1123  for (;;) Line 1480  for (;;)
1480        case OP_EXACT:        case OP_EXACT:
1481        case OP_UPTO:        case OP_UPTO:
1482        case OP_MINUPTO:        case OP_MINUPTO:
1483          case OP_POSUPTO:
1484        case OP_STAR:        case OP_STAR:
1485        case OP_MINSTAR:        case OP_MINSTAR:
1486          case OP_POSSTAR:
1487        case OP_PLUS:        case OP_PLUS:
1488        case OP_MINPLUS:        case OP_MINPLUS:
1489          case OP_POSPLUS:
1490        case OP_QUERY:        case OP_QUERY:
1491        case OP_MINQUERY:        case OP_MINQUERY:
1492        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1493        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;  
1494        break;        break;
1495        }        }
1496  #endif  #endif
# Line 1152  for (;;) Line 1505  for (;;)
1505  *************************************************/  *************************************************/
1506    
1507  /* 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
1508  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()
1509  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
1510  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
1511  whose current branch will already have been scanned.  assertions. If we hit an unclosed bracket, we return "empty" - this means we've
1512    struck an inner bracket whose current branch will already have been scanned.
1513    
1514  Arguments:  Arguments:
1515    code        points to start of search    code        points to start of search
# Line 1169  static BOOL Line 1523  static BOOL
1523  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1524  {  {
1525  register int c;  register int c;
1526  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);
1527       code < endcode;       code < endcode;
1528       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1529    {    {
# Line 1177  for (code = first_significant_code(code Line 1531  for (code = first_significant_code(code
1531    
1532    c = *code;    c = *code;
1533    
1534    if (c >= OP_BRA)    /* Groups with zero repeats can of course be empty; skip them. */
1535    
1536      if (c == OP_BRAZERO || c == OP_BRAMINZERO)
1537        {
1538        code += _pcre_OP_lengths[c];
1539        do code += GET(code, 1); while (*code == OP_ALT);
1540        c = *code;
1541        continue;
1542        }
1543    
1544      /* For other groups, scan the branches. */
1545    
1546      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1547      {      {
1548      BOOL empty_branch;      BOOL empty_branch;
1549      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1193  for (code = first_significant_code(code Line 1559  for (code = first_significant_code(code
1559        }        }
1560      while (*code == OP_ALT);      while (*code == OP_ALT);
1561      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
     code += 1 + LINK_SIZE;  
1562      c = *code;      c = *code;
1563        continue;
1564      }      }
1565    
1566    else switch (c)    /* Handle the other opcodes */
1567    
1568      switch (c)
1569      {      {
1570      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1571        cannot be represented just by a bit map. This includes negated single
1572        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1573        actual length is stored in the compiled code, so we must update "code"
1574        here. */
1575    
1576  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1577      case OP_XCLASS:      case OP_XCLASS:
1578      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1579      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1580  #endif  #endif
1581    
# Line 1253  for (code = first_significant_code(code Line 1625  for (code = first_significant_code(code
1625      case OP_NOT:      case OP_NOT:
1626      case OP_PLUS:      case OP_PLUS:
1627      case OP_MINPLUS:      case OP_MINPLUS:
1628        case OP_POSPLUS:
1629      case OP_EXACT:      case OP_EXACT:
1630      case OP_NOTPLUS:      case OP_NOTPLUS:
1631      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1632        case OP_NOTPOSPLUS:
1633      case OP_NOTEXACT:      case OP_NOTEXACT:
1634      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1635      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1636        case OP_TYPEPOSPLUS:
1637      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1638      return FALSE;      return FALSE;
1639    
1640        /* These are going to continue, as they may be empty, but we have to
1641        fudge the length for the \p and \P cases. */
1642    
1643        case OP_TYPESTAR:
1644        case OP_TYPEMINSTAR:
1645        case OP_TYPEPOSSTAR:
1646        case OP_TYPEQUERY:
1647        case OP_TYPEMINQUERY:
1648        case OP_TYPEPOSQUERY:
1649        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1650        break;
1651    
1652        /* Same for these */
1653    
1654        case OP_TYPEUPTO:
1655        case OP_TYPEMINUPTO:
1656        case OP_TYPEPOSUPTO:
1657        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1658        break;
1659    
1660      /* End of branch */      /* End of branch */
1661    
1662      case OP_KET:      case OP_KET:
# Line 1270  for (code = first_significant_code(code Line 1665  for (code = first_significant_code(code
1665      case OP_ALT:      case OP_ALT:
1666      return TRUE;      return TRUE;
1667    
1668      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1669      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1670    
1671  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1672      case OP_STAR:      case OP_STAR:
1673      case OP_MINSTAR:      case OP_MINSTAR:
1674        case OP_POSSTAR:
1675      case OP_QUERY:      case OP_QUERY:
1676      case OP_MINQUERY:      case OP_MINQUERY:
1677        case OP_POSQUERY:
1678      case OP_UPTO:      case OP_UPTO:
1679      case OP_MINUPTO:      case OP_MINUPTO:
1680        case OP_POSUPTO:
1681      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1682      break;      break;
1683  #endif  #endif
# Line 1375  Returns:     a value representing the na Line 1773  Returns:     a value representing the na
1773  static int  static int
1774  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1775  {  {
1776    const char *pn = posix_names;
1777  register int yield = 0;  register int yield = 0;
1778  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1779    {    {
1780    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1781      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1782      pn += posix_name_lengths[yield] + 1;
1783    yield++;    yield++;
1784    }    }
1785  return -1;  return -1;
# Line 1397  earlier groups that are outside the curr Line 1797  earlier groups that are outside the curr
1797  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
1798  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
1799  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
1800  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,
1801  partially compiled regex must be temporarily terminated with OP_END.  the partially compiled regex must be temporarily terminated with OP_END.
1802    
1803    This function has been extended with the possibility of forward references for
1804    recursions and subroutine calls. It must also check the list of such references
1805    for the group we are dealing with. If it finds that one of the recursions in
1806    the current group is on this list, it adjusts the offset in the list, not the
1807    value in the reference (which is a group number).
1808    
1809  Arguments:  Arguments:
1810    group      points to the start of the group    group      points to the start of the group
1811    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1812    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1813    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1814      save_hwm   the hwm forward reference pointer at the start of the group
1815    
1816  Returns:     nothing  Returns:     nothing
1817  */  */
1818    
1819  static void  static void
1820  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1821      uschar *save_hwm)
1822  {  {
1823  uschar *ptr = group;  uschar *ptr = group;
1824    
1825  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1826    {    {
1827    int offset = GET(ptr, 1);    int offset;
1828    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1829    
1830      /* See if this recursion is on the forward reference list. If so, adjust the
1831      reference. */
1832    
1833      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1834        {
1835        offset = GET(hc, 0);
1836        if (cd->start_code + offset == ptr + 1)
1837          {
1838          PUT(hc, 0, offset + adjust);
1839          break;
1840          }
1841        }
1842    
1843      /* Otherwise, adjust the recursion offset if it's after the start of this
1844      group. */
1845    
1846      if (hc >= cd->hwm)
1847        {
1848        offset = GET(ptr, 1);
1849        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1850        }
1851    
1852    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1853    }    }
1854  }  }
# Line 1495  Yield:        TRUE when range returned; Line 1927  Yield:        TRUE when range returned;
1927  */  */
1928    
1929  static BOOL  static BOOL
1930  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
1931      unsigned int *odptr)
1932  {  {
1933  int c, chartype, othercase, next;  unsigned int c, othercase, next;
1934    
1935  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
1936    {    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
1937    
1938  if (c > d) return FALSE;  if (c > d) return FALSE;
1939    
# Line 1512  next = othercase + 1; Line 1942  next = othercase + 1;
1942    
1943  for (++c; c <= d; c++)  for (++c; c <= d; c++)
1944    {    {
1945    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (_pcre_ucp_othercase(c) != next) break;
         othercase != next)  
     break;  
1946    next++;    next++;
1947    }    }
1948    
# Line 1526  return TRUE; Line 1954  return TRUE;
1954  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
1955    
1956    
1957    
1958  /*************************************************  /*************************************************
1959  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
1960  *************************************************/  *************************************************/
1961    
1962  /* 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
1963  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
1964  bits.  sense to automatically possessify the repeated item.
1965    
1966  Arguments:  Arguments:
1967    optionsptr     pointer to the option bits    op_code       the repeated op code
1968    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
1969    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
1970    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
1971    errorcodeptr   points to error code variable    ptr           next character in pattern
1972    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
1973    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.  
1974    
1975  Returns:         TRUE on success  Returns:        TRUE if possessifying is wanted
                  FALSE, with *errorcodeptr set non-zero on error  
1976  */  */
1977    
1978  static BOOL  static BOOL
1979  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
1980    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const uschar *ptr, int options, compile_data *cd)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
1981  {  {
1982  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  
1983    
1984  /* Set up the default and non-default settings for greediness */  /* Skip whitespace and comments in extended mode */
1985    
1986    if ((options & PCRE_EXTENDED) != 0)
1987      {
1988      for (;;)
1989        {
1990        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
1991        if (*ptr == '#')
1992          {
1993          while (*(++ptr) != 0)
1994            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
1995          }
1996        else break;
1997        }
1998      }
1999    
2000    /* If the next item is one that we can handle, get its value. A non-negative
2001    value is a character, a negative value is an escape value. */
2002    
2003    if (*ptr == '\\')
2004      {
2005      int temperrorcode = 0;
2006      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2007      if (temperrorcode != 0) return FALSE;
2008      ptr++;    /* Point after the escape sequence */
2009      }
2010    
2011    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2012      {
2013    #ifdef SUPPORT_UTF8
2014      if (utf8) { GETCHARINC(next, ptr); } else
2015    #endif
2016      next = *ptr++;
2017      }
2018    
2019    else return FALSE;
2020    
2021    /* Skip whitespace and comments in extended mode */
2022    
2023    if ((options & PCRE_EXTENDED) != 0)
2024      {
2025      for (;;)
2026        {
2027        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2028        if (*ptr == '#')
2029          {
2030          while (*(++ptr) != 0)
2031            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2032          }
2033        else break;
2034        }
2035      }
2036    
2037    /* If the next thing is itself optional, we have to give up. */
2038    
2039    if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
2040      return FALSE;
2041    
2042    /* Now compare the next item with the previous opcode. If the previous is a
2043    positive single character match, "item" either contains the character or, if
2044    "item" is greater than 127 in utf8 mode, the character's bytes are in
2045    utf8_char. */
2046    
2047    
2048    /* Handle cases when the next item is a character. */
2049    
2050    if (next >= 0) switch(op_code)
2051      {
2052      case OP_CHAR:
2053    #ifdef SUPPORT_UTF8
2054      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2055    #endif
2056      return item != next;
2057    
2058      /* For CHARNC (caseless character) we must check the other case. If we have
2059      Unicode property support, we can use it to test the other case of
2060      high-valued characters. */
2061    
2062      case OP_CHARNC:
2063    #ifdef SUPPORT_UTF8
2064      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2065    #endif
2066      if (item == next) return FALSE;
2067    #ifdef SUPPORT_UTF8
2068      if (utf8)
2069        {
2070        unsigned int othercase;
2071        if (next < 128) othercase = cd->fcc[next]; else
2072    #ifdef SUPPORT_UCP
2073        othercase = _pcre_ucp_othercase((unsigned int)next);
2074    #else
2075        othercase = NOTACHAR;
2076    #endif
2077        return (unsigned int)item != othercase;
2078        }
2079      else
2080    #endif  /* SUPPORT_UTF8 */
2081      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2082    
2083      /* For OP_NOT, "item" must be a single-byte character. */
2084    
2085      case OP_NOT:
2086      if (next < 0) return FALSE;  /* Not a character */
2087      if (item == next) return TRUE;
2088      if ((options & PCRE_CASELESS) == 0) return FALSE;
2089    #ifdef SUPPORT_UTF8
2090      if (utf8)
2091        {
2092        unsigned int othercase;
2093        if (next < 128) othercase = cd->fcc[next]; else
2094    #ifdef SUPPORT_UCP
2095        othercase = _pcre_ucp_othercase(next);
2096    #else
2097        othercase = NOTACHAR;
2098    #endif
2099        return (unsigned int)item == othercase;
2100        }
2101      else
2102    #endif  /* SUPPORT_UTF8 */
2103      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2104    
2105      case OP_DIGIT:
2106      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2107    
2108      case OP_NOT_DIGIT:
2109      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2110    
2111      case OP_WHITESPACE:
2112      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2113    
2114      case OP_NOT_WHITESPACE:
2115      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2116    
2117      case OP_WORDCHAR:
2118      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2119    
2120      case OP_NOT_WORDCHAR:
2121      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2122    
2123      case OP_HSPACE:
2124      case OP_NOT_HSPACE:
2125      switch(next)
2126        {
2127        case 0x09:
2128        case 0x20:
2129        case 0xa0:
2130        case 0x1680:
2131        case 0x180e:
2132        case 0x2000:
2133        case 0x2001:
2134        case 0x2002:
2135        case 0x2003:
2136        case 0x2004:
2137        case 0x2005:
2138        case 0x2006:
2139        case 0x2007:
2140        case 0x2008:
2141        case 0x2009:
2142        case 0x200A:
2143        case 0x202f:
2144        case 0x205f:
2145        case 0x3000:
2146        return op_code != OP_HSPACE;
2147        default:
2148        return op_code == OP_HSPACE;
2149        }
2150    
2151      case OP_VSPACE:
2152      case OP_NOT_VSPACE:
2153      switch(next)
2154        {
2155        case 0x0a:
2156        case 0x0b:
2157        case 0x0c:
2158        case 0x0d:
2159        case 0x85:
2160        case 0x2028:
2161        case 0x2029:
2162        return op_code != OP_VSPACE;
2163        default:
2164        return op_code == OP_VSPACE;
2165        }
2166    
2167      default:
2168      return FALSE;
2169      }
2170    
2171    
2172    /* Handle the case when the next item is \d, \s, etc. */
2173    
2174    switch(op_code)
2175      {
2176      case OP_CHAR:
2177      case OP_CHARNC:
2178    #ifdef SUPPORT_UTF8
2179      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2180    #endif
2181      switch(-next)
2182        {
2183        case ESC_d:
2184        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2185    
2186        case ESC_D:
2187        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2188    
2189        case ESC_s:
2190        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2191    
2192        case ESC_S:
2193        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2194    
2195        case ESC_w:
2196        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2197    
2198        case ESC_W:
2199        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2200    
2201        case ESC_h:
2202        case ESC_H:
2203        switch(item)
2204          {
2205          case 0x09:
2206          case 0x20:
2207          case 0xa0:
2208          case 0x1680:
2209          case 0x180e:
2210          case 0x2000:
2211          case 0x2001:
2212          case 0x2002:
2213          case 0x2003:
2214          case 0x2004:
2215          case 0x2005:
2216          case 0x2006:
2217          case 0x2007:
2218          case 0x2008:
2219          case 0x2009:
2220          case 0x200A:
2221          case 0x202f:
2222          case 0x205f:
2223          case 0x3000:
2224          return -next != ESC_h;
2225          default:
2226          return -next == ESC_h;
2227          }
2228    
2229        case ESC_v:
2230        case ESC_V:
2231        switch(item)
2232          {
2233          case 0x0a:
2234          case 0x0b:
2235          case 0x0c:
2236          case 0x0d:
2237          case 0x85:
2238          case 0x2028:
2239          case 0x2029:
2240          return -next != ESC_v;
2241          default:
2242          return -next == ESC_v;
2243          }
2244    
2245        default:
2246        return FALSE;
2247        }
2248    
2249      case OP_DIGIT:
2250      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2251             next == -ESC_h || next == -ESC_v;
2252    
2253      case OP_NOT_DIGIT:
2254      return next == -ESC_d;
2255    
2256      case OP_WHITESPACE:
2257      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2258    
2259      case OP_NOT_WHITESPACE:
2260      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2261    
2262      case OP_HSPACE:
2263      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2264    
2265      case OP_NOT_HSPACE:
2266      return next == -ESC_h;
2267    
2268      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2269      case OP_VSPACE:
2270      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2271    
2272      case OP_NOT_VSPACE:
2273      return next == -ESC_v;
2274    
2275      case OP_WORDCHAR:
2276      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2277    
2278      case OP_NOT_WORDCHAR:
2279      return next == -ESC_w || next == -ESC_d;
2280    
2281      default:
2282      return FALSE;
2283      }
2284    
2285    /* Control does not reach here */
2286    }
2287    
2288    
2289    
2290    /*************************************************
2291    *           Compile one branch                   *
2292    *************************************************/
2293    
2294    /* Scan the pattern, compiling it into the a vector. If the options are
2295    changed during the branch, the pointer is used to change the external options
2296    bits. This function is used during the pre-compile phase when we are trying
2297    to find out the amount of memory needed, as well as during the real compile
2298    phase. The value of lengthptr distinguishes the two phases.
2299    
2300    Arguments:
2301      optionsptr     pointer to the option bits
2302      codeptr        points to the pointer to the current code point
2303      ptrptr         points to the current pattern pointer
2304      errorcodeptr   points to error code variable
2305      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2306      reqbyteptr     set to the last literal character required, else < 0
2307      bcptr          points to current branch chain
2308      cd             contains pointers to tables etc.
2309      lengthptr      NULL during the real compile phase
2310                     points to length accumulator during pre-compile phase
2311    
2312    Returns:         TRUE on success
2313                     FALSE, with *errorcodeptr set non-zero on error
2314    */
2315    
2316    static BOOL
2317    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2318      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2319      compile_data *cd, int *lengthptr)
2320    {
2321    int repeat_type, op_type;
2322    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2323    int bravalue = 0;
2324    int greedy_default, greedy_non_default;
2325    int firstbyte, reqbyte;
2326    int zeroreqbyte, zerofirstbyte;
2327    int req_caseopt, reqvary, tempreqvary;
2328    int options = *optionsptr;
2329    int after_manual_callout = 0;
2330    int length_prevgroup = 0;
2331    register int c;
2332    register uschar *code = *codeptr;
2333    uschar *last_code = code;
2334    uschar *orig_code = code;
2335    uschar *tempcode;
2336    BOOL inescq = FALSE;
2337    BOOL groupsetfirstbyte = FALSE;
2338    const uschar *ptr = *ptrptr;
2339    const uschar *tempptr;
2340    uschar *previous = NULL;
2341    uschar *previous_callout = NULL;
2342    uschar *save_hwm = NULL;
2343    uschar classbits[32];
2344    
2345    #ifdef SUPPORT_UTF8
2346    BOOL class_utf8;
2347    BOOL utf8 = (options & PCRE_UTF8) != 0;
2348    uschar *class_utf8data;
2349    uschar utf8_char[6];
2350    #else
2351    BOOL utf8 = FALSE;
2352    uschar *utf8_char = NULL;
2353    #endif
2354    
2355    #ifdef DEBUG
2356    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2357    #endif
2358    
2359    /* Set up the default and non-default settings for greediness */
2360    
2361  greedy_default = ((options & PCRE_UNGREEDY) != 0);  greedy_default = ((options & PCRE_UNGREEDY) != 0);
2362  greedy_non_default = greedy_default ^ 1;  greedy_non_default = greedy_default ^ 1;
# Line 1613  req_caseopt = ((options & PCRE_CASELESS) Line 2385  req_caseopt = ((options & PCRE_CASELESS)
2385  for (;; ptr++)  for (;; ptr++)
2386    {    {
2387    BOOL negate_class;    BOOL negate_class;
2388      BOOL should_flip_negation;
2389    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2390    BOOL is_quantifier;    BOOL is_quantifier;
2391      BOOL is_recurse;
2392      BOOL reset_bracount;
2393    int class_charcount;    int class_charcount;
2394    int class_lastchar;    int class_lastchar;
2395    int newoptions;    int newoptions;
2396    int recno;    int recno;
2397      int refsign;
2398    int skipbytes;    int skipbytes;
2399    int subreqbyte;    int subreqbyte;
2400    int subfirstbyte;    int subfirstbyte;
2401      int terminator;
2402    int mclength;    int mclength;
2403    uschar mcbuffer[8];    uschar mcbuffer[8];
2404    
2405    /* Next byte in the pattern */    /* Get next byte in the pattern */
2406    
2407    c = *ptr;    c = *ptr;
2408    
2409      /* If we are in the pre-compile phase, accumulate the length used for the
2410      previous cycle of this loop. */
2411    
2412      if (lengthptr != NULL)
2413        {
2414    #ifdef DEBUG
2415        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2416    #endif
2417        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2418          {
2419          *errorcodeptr = ERR52;
2420          goto FAILED;
2421          }
2422    
2423        /* There is at least one situation where code goes backwards: this is the
2424        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2425        the class is simply eliminated. However, it is created first, so we have to
2426        allow memory for it. Therefore, don't ever reduce the length at this point.
2427        */
2428    
2429        if (code < last_code) code = last_code;
2430    
2431        /* Paranoid check for integer overflow */
2432    
2433        if (OFLOW_MAX - *lengthptr < code - last_code)
2434          {
2435          *errorcodeptr = ERR20;
2436          goto FAILED;
2437          }
2438    
2439        *lengthptr += code - last_code;
2440        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2441    
2442        /* If "previous" is set and it is not at the start of the work space, move
2443        it back to there, in order to avoid filling up the work space. Otherwise,
2444        if "previous" is NULL, reset the current code pointer to the start. */
2445    
2446        if (previous != NULL)
2447          {
2448          if (previous > orig_code)
2449            {
2450            memmove(orig_code, previous, code - previous);
2451            code -= previous - orig_code;
2452            previous = orig_code;
2453            }
2454          }
2455        else code = orig_code;
2456    
2457        /* Remember where this code item starts so we can pick up the length
2458        next time round. */
2459    
2460        last_code = code;
2461        }
2462    
2463      /* In the real compile phase, just check the workspace used by the forward
2464      reference list. */
2465    
2466      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2467        {
2468        *errorcodeptr = ERR52;
2469        goto FAILED;
2470        }
2471    
2472    /* 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 */
2473    
2474    if (inescq && c != 0)    if (inescq && c != 0)
# Line 1643  for (;; ptr++) Line 2483  for (;; ptr++)
2483        {        {
2484        if (previous_callout != NULL)        if (previous_callout != NULL)
2485          {          {
2486          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2487              complete_callout(previous_callout, ptr, cd);
2488          previous_callout = NULL;          previous_callout = NULL;
2489          }          }
2490        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1664  for (;; ptr++) Line 2505  for (;; ptr++)
2505    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2506         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2507      {      {
2508      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2509          complete_callout(previous_callout, ptr, cd);
2510      previous_callout = NULL;      previous_callout = NULL;
2511      }      }
2512    
# Line 1675  for (;; ptr++) Line 2517  for (;; ptr++)
2517      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2518      if (c == '#')      if (c == '#')
2519        {        {
2520        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2521        on the Macintosh. */          {
2522        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2523        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2524          if (*ptr != 0) continue;
2525    
2526          /* Else fall through to handle end of string */
2527          c = 0;
2528        }        }
2529      }      }
2530    
# Line 1692  for (;; ptr++) Line 2538  for (;; ptr++)
2538    
2539    switch(c)    switch(c)
2540      {      {
2541      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2542        case 0:                        /* The branch terminates at string end */
2543      case 0:      case '|':                      /* or | or ) */
     case '|':  
2544      case ')':      case ')':
2545      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2546      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2547      *codeptr = code;      *codeptr = code;
2548      *ptrptr = ptr;      *ptrptr = ptr;
2549        if (lengthptr != NULL)
2550          {
2551          if (OFLOW_MAX - *lengthptr < code - last_code)
2552            {
2553            *errorcodeptr = ERR20;
2554            goto FAILED;
2555            }
2556          *lengthptr += code - last_code;   /* To include callout length */
2557          DPRINTF((">> end branch\n"));
2558          }
2559      return TRUE;      return TRUE;
2560    
2561    
2562        /* ===================================================================*/
2563      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2564      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2565    
# Line 1731  for (;; ptr++) Line 2588  for (;; ptr++)
2588      *code++ = OP_ANY;      *code++ = OP_ANY;
2589      break;      break;
2590    
2591      /* Character classes. If the included characters are all < 255 in value, we  
2592      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2593      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
2594      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
2595      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2596        map as usual, then invert it at the end. However, we use a different opcode
2597        so that data characters > 255 can be handled correctly.
2598    
2599      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2600      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 1756  for (;; ptr++) Line 2615  for (;; ptr++)
2615        goto FAILED;        goto FAILED;
2616        }        }
2617    
2618      /* 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,
2619        if the first few characters (either before or after ^) are \Q\E or \E we
2620        skip them too. This makes for compatibility with Perl. */
2621    
2622      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2623        for (;;)
2624        {        {
       negate_class = TRUE;  
2625        c = *(++ptr);        c = *(++ptr);
2626          if (c == '\\')
2627            {
2628            if (ptr[1] == 'E') ptr++;
2629              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2630                else break;
2631            }
2632          else if (!negate_class && c == '^')
2633            negate_class = TRUE;
2634          else break;
2635        }        }
2636      else  
2637        {      /* If a class contains a negative special such as \S, we need to flip the
2638        negate_class = FALSE;      negation flag at the end, so that support for characters > 255 works
2639        }      correctly (they are all included in the class). */
2640    
2641        should_flip_negation = FALSE;
2642    
2643      /* 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
2644      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
2645      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2646    
2647      class_charcount = 0;      class_charcount = 0;
2648      class_lastchar = -1;      class_lastchar = -1;
2649    
2650        /* Initialize the 32-char bit map to all zeros. We build the map in a
2651        temporary bit of memory, in case the class contains only 1 character (less
2652        than 256), because in that case the compiled code doesn't use the bit map.
2653        */
2654    
2655        memset(classbits, 0, 32 * sizeof(uschar));
2656    
2657  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2658      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2659      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2660  #endif  #endif
2661    
     /* 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));  
   
2662      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2663      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
2664      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. */  
2665    
2666      do      if (c != 0) do
2667        {        {
2668          const uschar *oldptr;
2669    
2670  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2671        if (utf8 && c > 127)        if (utf8 && c > 127)
2672          {                           /* Braces are required because the */          {                           /* Braces are required because the */
# Line 1806  for (;; ptr++) Line 2678  for (;; ptr++)
2678    
2679        if (inescq)        if (inescq)
2680          {          {
2681          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2682            {            {
2683            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2684            ptr++;            ptr++;                            /* Skip the 'E' */
2685            continue;            continue;                         /* Carry on with next */
2686            }            }
2687          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2688          }          }
2689    
2690        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1826  for (;; ptr++) Line 2698  for (;; ptr++)
2698            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr, cd))
2699          {          {
2700          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2701          int posix_class, i;          int posix_class, taboffset, tabopt;
2702          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2703            uschar pbits[32];
2704    
2705          if (ptr[1] != ':')          if (ptr[1] != ':')
2706            {            {
# Line 1839  for (;; ptr++) Line 2712  for (;; ptr++)
2712          if (*ptr == '^')          if (*ptr == '^')
2713            {            {
2714            local_negate = TRUE;            local_negate = TRUE;
2715              should_flip_negation = TRUE;  /* Note negative special */
2716            ptr++;            ptr++;
2717            }            }
2718    
# Line 1856  for (;; ptr++) Line 2730  for (;; ptr++)
2730          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
2731            posix_class = 0;            posix_class = 0;
2732    
2733          /* 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
2734          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
2735          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
2736          white space chars afterwards. */          result into the bit map that is being built. */
2737    
2738          posix_class *= 3;          posix_class *= 3;
2739          for (i = 0; i < 3; i++)  
2740            /* Copy in the first table (always present) */
2741    
2742            memcpy(pbits, cbits + posix_class_maps[posix_class],
2743              32 * sizeof(uschar));
2744    
2745            /* If there is a second table, add or remove it as required. */
2746    
2747            taboffset = posix_class_maps[posix_class + 1];
2748            tabopt = posix_class_maps[posix_class + 2];
2749    
2750            if (taboffset >= 0)
2751            {            {
2752            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
2753            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;  
             }  
2754            else            else
2755              {              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;  
             }  
2756            }            }
2757    
2758            /* Not see if we need to remove any special characters. An option
2759            value of 1 removes vertical space and 2 removes underscore. */
2760    
2761            if (tabopt < 0) tabopt = -tabopt;
2762            if (tabopt == 1) pbits[1] &= ~0x3c;
2763              else if (tabopt == 2) pbits[11] &= 0x7f;
2764    
2765            /* Add the POSIX table or its complement into the main table that is
2766            being built and we are done. */
2767    
2768            if (local_negate)
2769              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
2770            else
2771              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
2772    
2773          ptr = tempptr + 1;          ptr = tempptr + 1;
2774          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
2775          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
2776          }          }
2777    
2778        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2779        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
2780        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.
2781        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2782        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  
2783        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2784    
2785        if (c == '\\')        if (c == '\\')
2786          {          {
2787          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2788            if (*errorcodeptr != 0) goto FAILED;
2789    
2790          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2791          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 */
2792            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2793          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2794            {            {
2795            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1910  for (;; ptr++) Line 2799  for (;; ptr++)
2799            else inescq = TRUE;            else inescq = TRUE;
2800            continue;            continue;
2801            }            }
2802            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2803    
2804          if (c < 0)          if (c < 0)
2805            {            {
2806            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2807            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2808            switch (-c)  
2809              /* Save time by not doing this in the pre-compile phase. */
2810    
2811              if (lengthptr == NULL) switch (-c)
2812              {              {
2813              case ESC_d:              case ESC_d:
2814              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
2815              continue;              continue;
2816    
2817              case ESC_D:              case ESC_D:
2818                should_flip_negation = TRUE;
2819              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2820              continue;              continue;
2821    
# Line 1930  for (;; ptr++) Line 2824  for (;; ptr++)
2824              continue;              continue;
2825    
2826              case ESC_W:              case ESC_W:
2827                should_flip_negation = TRUE;
2828              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2829              continue;              continue;
2830    
# Line 1939  for (;; ptr++) Line 2834  for (;; ptr++)
2834              continue;              continue;
2835    
2836              case ESC_S:              case ESC_S:
2837                should_flip_negation = TRUE;
2838              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2839              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2840              continue;              continue;
2841    
2842  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
2843              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
2844              case ESC_P:              }
2845    
2846              /* In the pre-compile phase, just do the recognition. */
2847    
2848              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2849                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
2850    
2851              /* We need to deal with \H, \h, \V, and \v in both phases because
2852              they use extra memory. */
2853    
2854              if (-c == ESC_h)
2855                {
2856                SETBIT(classbits, 0x09); /* VT */
2857                SETBIT(classbits, 0x20); /* SPACE */
2858                SETBIT(classbits, 0xa0); /* NSBP */
2859    #ifdef SUPPORT_UTF8
2860                if (utf8)
2861                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
2862                class_utf8 = TRUE;                class_utf8 = TRUE;
2863                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_SINGLE;
2864                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2865                *class_utf8data++ = property;                *class_utf8data++ = XCL_SINGLE;
2866                class_charcount -= 2;   /* Not a < 256 character */                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2867                  *class_utf8data++ = XCL_RANGE;
2868                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2869                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2870                  *class_utf8data++ = XCL_SINGLE;
2871                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2872                  *class_utf8data++ = XCL_SINGLE;
2873                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2874                  *class_utf8data++ = XCL_SINGLE;
2875                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2876                }                }
             continue;  
2877  #endif  #endif
2878                continue;
2879                }
2880    
2881              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_H)
2882              strict mode. By default, for compatibility with Perl, they are              {
2883              treated as literals. */              for (c = 0; c < 32; c++)
   
             default:  
             if ((options & PCRE_EXTRA) != 0)  
2884                {                {
2885                *errorcodeptr = ERR7;                int x = 0xff;
2886                goto FAILED;                switch (c)
2887                    {
2888                    case 0x09/8: x ^= 1 << (0x09%8); break;
2889                    case 0x20/8: x ^= 1 << (0x20%8); break;
2890                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2891                    default: break;
2892                    }
2893                  classbits[c] |= x;
2894                }                }
2895              c = *ptr;              /* The final character */  
2896              class_charcount -= 2;  /* Undo the default count from above */  #ifdef SUPPORT_UTF8
2897                if (utf8)
2898                  {
2899                  class_utf8 = TRUE;
2900                  *class_utf8data++ = XCL_RANGE;
2901                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2902                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
2903                  *class_utf8data++ = XCL_RANGE;
2904                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
2905                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
2906                  *class_utf8data++ = XCL_RANGE;
2907                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
2908                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
2909                  *class_utf8data++ = XCL_RANGE;
2910                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
2911                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
2912                  *class_utf8data++ = XCL_RANGE;
2913                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
2914                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
2915                  *class_utf8data++ = XCL_RANGE;
2916                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
2917                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
2918                  *class_utf8data++ = XCL_RANGE;
2919                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
2920                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2921                  }
2922    #endif
2923                continue;
2924                }
2925    
2926              if (-c == ESC_v)
2927                {
2928                SETBIT(classbits, 0x0a); /* LF */
2929                SETBIT(classbits, 0x0b); /* VT */
2930                SETBIT(classbits, 0x0c); /* FF */
2931                SETBIT(classbits, 0x0d); /* CR */
2932                SETBIT(classbits, 0x85); /* NEL */
2933    #ifdef SUPPORT_UTF8
2934                if (utf8)
2935                  {
2936                  class_utf8 = TRUE;
2937                  *class_utf8data++ = XCL_RANGE;
2938                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
2939                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2940                  }
2941    #endif
2942                continue;
2943                }
2944    
2945              if (-c == ESC_V)
2946                {
2947                for (c = 0; c < 32; c++)
2948                  {
2949                  int x = 0xff;
2950                  switch (c)
2951                    {
2952                    case 0x0a/8: x ^= 1 << (0x0a%8);
2953                                 x ^= 1 << (0x0b%8);
2954                                 x ^= 1 << (0x0c%8);
2955                                 x ^= 1 << (0x0d%8);
2956                                 break;
2957                    case 0x85/8: x ^= 1 << (0x85%8); break;
2958                    default: break;
2959                    }
2960                  classbits[c] |= x;
2961                  }
2962    
2963    #ifdef SUPPORT_UTF8
2964                if (utf8)
2965                  {
2966                  class_utf8 = TRUE;
2967                  *class_utf8data++ = XCL_RANGE;
2968                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2969                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
2970                  *class_utf8data++ = XCL_RANGE;
2971                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2972                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2973                  }
2974    #endif
2975                continue;
2976                }
2977    
2978              /* We need to deal with \P and \p in both phases. */
2979    
2980    #ifdef SUPPORT_UCP
2981              if (-c == ESC_p || -c == ESC_P)
2982                {
2983                BOOL negated;
2984                int pdata;
2985                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
2986                if (ptype < 0) goto FAILED;
2987                class_utf8 = TRUE;
2988                *class_utf8data++ = ((-c == ESC_p) != negated)?
2989                  XCL_PROP : XCL_NOTPROP;
2990                *class_utf8data++ = ptype;
2991                *class_utf8data++ = pdata;
2992                class_charcount -= 2;   /* Not a < 256 character */
2993                continue;
2994              }              }
2995    #endif
2996              /* Unrecognized escapes are faulted if PCRE is running in its
2997              strict mode. By default, for compatibility with Perl, they are
2998              treated as literals. */
2999    
3000              if ((options & PCRE_EXTRA) != 0)
3001                {
3002                *errorcodeptr = ERR7;
3003                goto FAILED;
3004                }
3005    
3006              class_charcount -= 2;  /* Undo the default count from above */
3007              c = *ptr;              /* Get the final character and fall through */
3008            }            }
3009    
3010          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
3011          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3012    
3013          }   /* End of backslash handling */          }   /* End of backslash handling */
3014    
3015        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3016        Perl does not permit ']' to be the end of the range. A '-' character        Perl does not permit ']' to be the end of the range. A '-' character
3017        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3018          entirely. The code for handling \Q and \E is messy. */
3019    
3020          CHECK_RANGE:
3021          while (ptr[1] == '\\' && ptr[2] == 'E')
3022            {
3023            inescq = FALSE;
3024            ptr += 2;
3025            }
3026    
3027          oldptr = ptr;
3028    
3029          /* Remember \r or \n */
3030    
3031          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3032    
3033          /* Check for range */
3034    
3035        if (ptr[1] == '-' && ptr[2] != ']')        if (!inescq && ptr[1] == '-')
3036          {          {
3037          int d;          int d;
3038          ptr += 2;          ptr += 2;
3039            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
3040    
3041            /* If we hit \Q (not followed by \E) at this point, go into escaped
3042            mode. */
3043    
3044            while (*ptr == '\\' && ptr[1] == 'Q')
3045              {
3046              ptr += 2;
3047              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
3048              inescq = TRUE;
3049              break;
3050              }
3051    
3052            if (*ptr == 0 || (!inescq && *ptr == ']'))
3053              {
3054              ptr = oldptr;
3055              goto LONE_SINGLE_CHARACTER;
3056              }
3057    
3058  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3059          if (utf8)          if (utf8)
# Line 2001  for (;; ptr++) Line 3068  for (;; ptr++)
3068          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
3069          in such circumstances. */          in such circumstances. */
3070    
3071          if (d == '\\')          if (!inescq && d == '\\')
3072            {            {
3073            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3074            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3075    
3076            /* \b is backslash; \X is literal X; any other special means the '-'            /* \b is backspace; \X is literal X; \R is literal R; any other
3077            was literal */            special means the '-' was literal */
3078    
3079            if (d < 0)            if (d < 0)
3080              {              {
3081              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = '\b';
3082              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = 'X';
3083                else if (d == -ESC_R) d = 'R'; else
3084                {                {
3085                ptr = oldptr - 2;                ptr = oldptr;
3086                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3087                }                }
3088              }              }
3089            }            }
3090    
3091          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3092          the pre-pass. Optimize one-character ranges */          one-character ranges */
3093    
3094            if (d < c)
3095              {
3096              *errorcodeptr = ERR8;
3097              goto FAILED;
3098              }
3099    
3100          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3101    
3102            /* Remember \r or \n */
3103    
3104            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3105    
3106          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
3107          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3108          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2042  for (;; ptr++) Line 3120  for (;; ptr++)
3120  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3121            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3122              {              {
3123              int occ, ocd;              unsigned int occ, ocd;
3124              int cc = c;              unsigned int cc = c;
3125              int origd = d;              unsigned int origd = d;
3126              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3127                {                {
3128                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3129                      ocd <= (unsigned int)d)
3130                    continue;                          /* Skip embedded ranges */
3131    
3132                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3133                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3134                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3135                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3136                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3137                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3138                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3139                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3140                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3141                  d = ocd;                  d = ocd;
3142                  continue;                  continue;
# Line 2102  for (;; ptr++) Line 3184  for (;; ptr++)
3184          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
3185          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3186    
3187          for (; c <= d; c++)          class_charcount += d - c + 1;
3188            class_lastchar = d;
3189    
3190            /* We can save a bit of time by skipping this in the pre-compile. */
3191    
3192            if (lengthptr == NULL) for (; c <= d; c++)
3193            {            {
3194            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3195            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2110  for (;; ptr++) Line 3197  for (;; ptr++)
3197              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3198              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3199              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3200            }            }
3201    
3202          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2135  for (;; ptr++) Line 3220  for (;; ptr++)
3220  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3221          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3222            {            {
3223            int chartype;            unsigned int othercase;
3224            int othercase;            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3225              {              {
3226              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3227              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2163  for (;; ptr++) Line 3246  for (;; ptr++)
3246          }          }
3247        }        }
3248    
3249      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3250      loop. This "while" is the end of the "do" above. */  
3251        while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
3252    
3253        if (c == 0)                          /* Missing terminating ']' */
3254          {
3255          *errorcodeptr = ERR6;
3256          goto FAILED;
3257          }
3258    
3259    
3260    /* This code has been disabled because it would mean that \s counts as
3261    an explicit \r or \n reference, and that's not really what is wanted. Now
3262    we set the flag only if there is a literal "\r" or "\n" in the class. */
3263    
3264    #if 0
3265        /* Remember whether \r or \n are in this class */
3266    
3267        if (negate_class)
3268          {
3269          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3270          }
3271        else
3272          {
3273          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3274          }
3275    #endif
3276    
     while ((c = *(++ptr)) != ']' || inescq);  
3277    
3278      /* 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
3279      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
3280      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
3281      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3282      single-bytes only. This is an historical hangover. Maybe one day we can  
3283      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3284        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3285        operate on single-bytes only. This is an historical hangover. Maybe one day
3286        we can tidy these opcodes to handle multi-byte characters.
3287    
3288      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
3289      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 2183  for (;; ptr++) Line 3293  for (;; ptr++)
3293      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3294    
3295  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3296      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3297            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3298  #else  #else
3299      if (class_charcount == 1)      if (class_charcount == 1)
3300  #endif  #endif
# Line 2229  for (;; ptr++) Line 3337  for (;; ptr++)
3337      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3338    
3339      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3340      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3341      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3342        the class, so any that were explicitly given as well can be ignored. If
3343        (when there are explicit characters > 255 that must be listed) there are no
3344        characters < 256, we can omit the bitmap in the actual compiled code. */
3345    
3346  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3347      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3348        {        {
3349        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3350        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3351        code += LINK_SIZE;        code += LINK_SIZE;
3352        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3353    
3354        /* 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;
3355        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3356    
3357        if (class_charcount > 0)        if (class_charcount > 0)
3358          {          {
3359          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3360            memmove(code + 32, code, class_utf8data - code);
3361          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3362          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;  
3363          }          }
3364          else code = class_utf8data;
3365    
3366        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3367    
# Line 2266  for (;; ptr++) Line 3370  for (;; ptr++)
3370        }        }
3371  #endif  #endif
3372    
3373      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, set the opcode to OP_CLASS or
3374      and copy it into the code vector. If this is the first thing in the branch,      OP_NCLASS, depending on whether the whole class was negated and whether
3375      there can be no first char setting, whatever the repeat count. Any reqbyte      there were negative specials such as \S in the class. Then copy the 32-byte
3376      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3377    
3378        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3379      if (negate_class)      if (negate_class)
3380        {        {
3381        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3382        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3383        }        }
3384      else      else
3385        {        {
       *code++ = OP_CLASS;  
3386        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3387        }        }
3388      code += 32;      code += 32;
3389      break;      break;
3390    
3391    
3392        /* ===================================================================*/
3393      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3394      has been tested above. */      has been tested above. */
3395    
# Line 2351  for (;; ptr++) Line 3457  for (;; ptr++)
3457        }        }
3458      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3459    
     /* 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;  
       }  
   
3460      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3461      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
3462      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 2398  for (;; ptr++) Line 3490  for (;; ptr++)
3490          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3491          }          }
3492    
3493          /* If the repetition is unlimited, it pays to see if the next thing on
3494          the line is something that cannot possibly match this character. If so,
3495          automatically possessifying this item gains some performance in the case
3496          where the match fails. */
3497    
3498          if (!possessive_quantifier &&
3499              repeat_max < 0 &&
3500              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3501                options, cd))
3502            {
3503            repeat_type = 0;    /* Force greedy */
3504            possessive_quantifier = TRUE;
3505            }
3506    
3507        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3508        }        }
3509    
3510      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3511      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-
3512      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3513      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3514        currently used only for single-byte chars. */
3515    
3516      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3517        {        {
3518        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3519        c = previous[1];        c = previous[1];
3520          if (!possessive_quantifier &&
3521              repeat_max < 0 &&
3522              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3523            {
3524            repeat_type = 0;    /* Force greedy */
3525            possessive_quantifier = TRUE;
3526            }
3527        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3528        }        }
3529    
# Line 2423  for (;; ptr++) Line 3537  for (;; ptr++)
3537      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3538        {        {
3539        uschar *oldcode;        uschar *oldcode;
3540        int prop_type;        int prop_type, prop_value;
3541        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3542        c = *previous;        c = *previous;
3543    
3544          if (!possessive_quantifier &&
3545              repeat_max < 0 &&
3546              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3547            {
3548            repeat_type = 0;    /* Force greedy */
3549            possessive_quantifier = TRUE;
3550            }
3551    
3552        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3553        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3554          previous[1] : -1;          {
3555            prop_type = previous[1];
3556            prop_value = previous[2];
3557            }
3558          else prop_type = prop_value = -1;
3559    
3560        oldcode = code;        oldcode = code;
3561        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2442  for (;; ptr++) Line 3568  for (;; ptr++)
3568        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3569        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3570    
3571        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3572    
3573        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3574    
# Line 2463  for (;; ptr++) Line 3589  for (;; ptr++)
3589          }          }
3590    
3591        /* 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
3592        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3593        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
3594        one less than the maximum. */        one less than the maximum. */
3595    
# Line 2490  for (;; ptr++) Line 3616  for (;; ptr++)
3616    
3617          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3618          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
3619          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3620          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3621          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3622    
# Line 2506  for (;; ptr++) Line 3632  for (;; ptr++)
3632  #endif  #endif
3633              {              {
3634              *code++ = c;              *code++ = c;
3635              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3636                  {
3637                  *code++ = prop_type;
3638                  *code++ = prop_value;
3639                  }
3640              }              }
3641            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3642            }            }
3643    
3644          /* 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
3645          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3646            UPTO is just for 1 instance, we can use QUERY instead. */
3647    
3648          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3649            {            {
# Line 2525  for (;; ptr++) Line 3656  for (;; ptr++)
3656            else            else
3657  #endif  #endif
3658            *code++ = c;            *code++ = c;
3659            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3660                {
3661                *code++ = prop_type;
3662                *code++ = prop_value;
3663                }
3664            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3665            *code++ = OP_UPTO + repeat_type;  
3666            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3667                {
3668                *code++ = OP_QUERY + repeat_type;
3669                }
3670              else
3671                {
3672                *code++ = OP_UPTO + repeat_type;
3673                PUT2INC(code, 0, repeat_max);
3674                }
3675            }            }
3676          }          }
3677    
# Line 2544  for (;; ptr++) Line 3687  for (;; ptr++)
3687  #endif  #endif
3688        *code++ = c;        *code++ = c;
3689    
3690        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3691        defines the required property. */        define the required property. */
3692    
3693  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3694        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3695            {
3696            *code++ = prop_type;
3697            *code++ = prop_value;
3698            }
3699  #endif  #endif
3700        }        }
3701    
# Line 2571  for (;; ptr++) Line 3718  for (;; ptr++)
3718        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3719        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3720    
3721        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3722    
3723        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3724          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2591  for (;; ptr++) Line 3738  for (;; ptr++)
3738      /* 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
3739      cases. */      cases. */
3740    
3741      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3742               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3743        {        {
3744        register int i;        register int i;
3745        int ketoffset = 0;        int ketoffset = 0;
3746        int len = code - previous;        int len = code - previous;
3747        uschar *bralink = NULL;        uschar *bralink = NULL;
3748    
3749          /* Repeating a DEFINE group is pointless */
3750    
3751          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3752            {
3753            *errorcodeptr = ERR55;
3754            goto FAILED;
3755            }
3756    
3757        /* 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
3758        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
3759        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 2633  for (;; ptr++) Line 3788  for (;; ptr++)
3788          /* 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
3789          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
3790          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
3791          any internal group, because the offset is from the start of the whole          any internal or forward referenced group, because the offset is from
3792          regex. Temporarily terminate the pattern while doing this. */          the start of the whole regex. Temporarily terminate the pattern while
3793            doing this. */
3794    
3795          if (repeat_max <= 1)          if (repeat_max <= 1)
3796            {            {
3797            *code = OP_END;            *code = OP_END;
3798            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3799            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3800            code++;            code++;
3801            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 3813  for (;; ptr++)
3813            {            {
3814            int offset;            int offset;
3815            *code = OP_END;            *code = OP_END;
3816            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3817            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3818            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3819            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2677  for (;; ptr++) Line 3833  for (;; ptr++)
3833        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3834        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3835        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
3836        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3837          forward reference subroutine calls in the group, there will be entries on
3838          the workspace list; replicate these with an appropriate increment. */
3839    
3840        else        else
3841          {          {
3842          if (repeat_min > 1)          if (repeat_min > 1)
3843            {            {
3844            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3845            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
3846              potential integer overflow. */
3847    
3848              if (lengthptr != NULL)
3849                {
3850                int delta = (repeat_min - 1)*length_prevgroup;
3851                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3852                                                                (double)INT_MAX ||
3853                    OFLOW_MAX - *lengthptr < delta)
3854                  {
3855                  *errorcodeptr = ERR20;
3856                  goto FAILED;
3857                  }
3858                *lengthptr += delta;
3859                }
3860    
3861              /* This is compiling for real */
3862    
3863              else
3864              {              {
3865              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3866              code += len;              for (i = 1; i < repeat_min; i++)
3867                  {
3868                  uschar *hc;
3869                  uschar *this_hwm = cd->hwm;
3870                  memcpy(code, previous, len);
3871                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3872                    {
3873                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3874                    cd->hwm += LINK_SIZE;
3875                    }
3876                  save_hwm = this_hwm;
3877                  code += len;
3878                  }
3879              }              }
3880            }            }
3881    
3882          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3883          }          }
3884    
# Line 2697  for (;; ptr++) Line 3886  for (;; ptr++)
3886        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3887        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,
3888        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
3889        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
3890          replicate entries on the forward reference list. */
3891    
3892        if (repeat_max >= 0)        if (repeat_max >= 0)
3893          {          {
3894          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
3895            just adjust the length as if we had. For each repetition we must add 1
3896            to the length for BRAZERO and for all but the last repetition we must
3897            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3898            paranoid checks to avoid integer overflow. */
3899    
3900            if (lengthptr != NULL && repeat_max > 0)
3901              {
3902              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3903                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3904              if ((double)repeat_max *
3905                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3906                      > (double)INT_MAX ||
3907                  OFLOW_MAX - *lengthptr < delta)
3908                {
3909                *errorcodeptr = ERR20;
3910                goto FAILED;
3911                }
3912              *lengthptr += delta;
3913              }
3914    
3915            /* This is compiling for real */
3916    
3917            else for (i = repeat_max - 1; i >= 0; i--)
3918            {            {
3919              uschar *hc;
3920              uschar *this_hwm = cd->hwm;
3921    
3922            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
3923    
3924            /* All but the final copy start a new nesting, maintaining the            /* All but the final copy start a new nesting, maintaining the
# Line 2718  for (;; ptr++) Line 3934  for (;; ptr++)
3934              }              }
3935    
3936            memcpy(code, previous, len);            memcpy(code, previous, len);
3937              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3938                {
3939                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
3940                cd->hwm += LINK_SIZE;
3941                }
3942              save_hwm = this_hwm;
3943            code += len;            code += len;
3944            }            }
3945    
# Line 2740  for (;; ptr++) Line 3962  for (;; ptr++)
3962        /* 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
3963        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
3964        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
3965        correct offset was computed above. */        correct offset was computed above.
3966    
3967        else code[-ketoffset] = OP_KETRMAX + repeat_type;        Then, when we are doing the actual compile phase, check to see whether
3968          this group is a non-atomic one that could match an empty string. If so,
3969          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
3970          that runtime checking can be done. [This check is also applied to
3971          atomic groups at runtime, but in a different way.] */
3972    
3973          else
3974            {
3975            uschar *ketcode = code - ketoffset;
3976            uschar *bracode = ketcode - GET(ketcode, 1);
3977            *ketcode = OP_KETRMAX + repeat_type;
3978            if (lengthptr == NULL && *bracode != OP_ONCE)
3979              {
3980              uschar *scode = bracode;
3981              do
3982                {
3983                if (could_be_empty_branch(scode, ketcode, utf8))
3984                  {
3985                  *bracode += OP_SBRA - OP_BRA;
3986                  break;
3987                  }
3988                scode += GET(scode, 1);
3989                }
3990              while (*scode == OP_ALT);
3991              }
3992            }
3993        }        }
3994    
3995      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2753  for (;; ptr++) Line 4000  for (;; ptr++)
4000        goto FAILED;        goto FAILED;
4001        }        }
4002    
4003      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
4004      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
4005      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
4006      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.
4007      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
4008        but the special opcodes can optimize it a bit. The repeated item starts at
4009        tempcode, not at previous, which might be the first part of a string whose
4010        (former) last char we repeated.
4011    
4012        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4013        an 'upto' may follow. We skip over an 'exact' item, and then test the
4014        length of what remains before proceeding. */
4015    
4016      if (possessive_quantifier)      if (possessive_quantifier)
4017        {        {
4018        int len = code - tempcode;        int len;
4019        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4020        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
4021        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode];
4022        tempcode[0] = OP_ONCE;        len = code - tempcode;
4023        *code++ = OP_KET;        if (len > 0) switch (*tempcode)
4024        PUTINC(code, 0, len);          {
4025        PUT(tempcode, 1, len);          case OP_STAR:  *tempcode = OP_POSSTAR; break;
4026            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
4027            case OP_QUERY: *tempcode = OP_POSQUERY; break;
4028            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4029    
4030            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4031            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4032            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4033            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4034    
4035            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4036            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4037            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4038            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4039    
4040            default:
4041            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4042            code += 1 + LINK_SIZE;
4043            len += 1 + LINK_SIZE;
4044            tempcode[0] = OP_ONCE;
4045            *code++ = OP_KET;
4046            PUTINC(code, 0, len);
4047            PUT(tempcode, 1, len);
4048            break;
4049            }
4050        }        }
4051    
4052      /* 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 2781  for (;; ptr++) Line 4059  for (;; ptr++)
4059      break;      break;
4060    
4061    
4062      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
4063      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4064      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
4065      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.  */  
4066    
4067      case '(':      case '(':
4068      newoptions = options;      newoptions = options;
4069      skipbytes = 0;      skipbytes = 0;
4070        bravalue = OP_CBRA;
4071        save_hwm = cd->hwm;
4072        reset_bracount = FALSE;
4073    
4074        /* First deal with various "verbs" that can be introduced by '*'. */
4075    
4076        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4077          {
4078          int i, namelen;
4079          const char *vn = verbnames;
4080          const uschar *name = ++ptr;
4081          previous = NULL;
4082          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4083          if (*ptr == ':')
4084            {
4085            *errorcodeptr = ERR59;   /* Not supported */
4086            goto FAILED;
4087            }
4088          if (*ptr != ')')
4089            {
4090            *errorcodeptr = ERR60;
4091            goto FAILED;
4092            }
4093          namelen = ptr - name;
4094          for (i = 0; i < verbcount; i++)
4095            {
4096            if (namelen == verbs[i].len &&
4097                strncmp((char *)name, vn, namelen) == 0)
4098              {
4099              *code = verbs[i].op;
4100              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4101              break;
4102              }
4103            vn += verbs[i].len + 1;
4104            }
4105          if (i < verbcount) continue;
4106          *errorcodeptr = ERR60;
4107          goto FAILED;
4108          }
4109    
4110        /* Deal with the extended parentheses; all are introduced by '?', and the
4111        appearance of any of them means that this is not a capturing group. */
4112    
4113      if (*(++ptr) == '?')      else if (*ptr == '?')
4114        {        {
4115        int set, unset;        int i, set, unset, namelen;
4116        int *optset;        int *optset;
4117          const uschar *name;
4118          uschar *slot;
4119    
4120        switch (*(++ptr))        switch (*(++ptr))
4121          {          {
4122          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
4123          ptr++;          ptr++;
4124          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
4125            if (*ptr == 0)
4126              {
4127              *errorcodeptr = ERR18;
4128              goto FAILED;
4129              }
4130          continue;          continue;
4131    
4132          case ':':                 /* Non-extracting bracket */  
4133            /* ------------------------------------------------------------ */
4134            case '|':                 /* Reset capture count for each branch */
4135            reset_bracount = TRUE;
4136            /* Fall through */
4137    
4138            /* ------------------------------------------------------------ */
4139            case ':':                 /* Non-capturing bracket */
4140          bravalue = OP_BRA;          bravalue = OP_BRA;
4141          ptr++;          ptr++;
4142          break;          break;
4143    
4144    
4145            /* ------------------------------------------------------------ */
4146          case '(':          case '(':
4147          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4148    
4149          /* Condition to test for recursion */          /* A condition can be an assertion, a number (referring to a numbered
4150            group), a name (referring to a named group), or 'R', referring to
4151            recursion. R<digits> and R&name are also permitted for recursion tests.
4152    
4153            There are several syntaxes for testing a named group: (?(name)) is used
4154            by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
4155    
4156            There are two unfortunate ambiguities, caused by history. (a) 'R' can
4157            be the recursive thing or the name 'R' (and similarly for 'R' followed
4158            by digits), and (b) a number could be a name that consists of digits.
4159            In both cases, we look for a name first; if not found, we try the other
4160            cases. */
4161    
4162            /* For conditions that are assertions, check the syntax, and then exit
4163            the switch. This will take control down to where bracketed groups,
4164            including assertions, are processed. */
4165    
4166            if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
4167              break;
4168    
4169            /* Most other conditions use OP_CREF (a couple change to OP_RREF
4170            below), and all need to skip 3 bytes at the start of the group. */
4171    
4172          if (ptr[1] == 'R')          code[1+LINK_SIZE] = OP_CREF;
4173            skipbytes = 3;
4174            refsign = -1;
4175    
4176            /* Check for a test for recursion in a named group. */
4177    
4178            if (ptr[1] == 'R' && ptr[2] == '&')
4179            {            {
4180            code[1+LINK_SIZE] = OP_CREF;            terminator = -1;
4181            PUT2(code, 2+LINK_SIZE, CREF_RECURSE);            ptr += 2;
4182            skipbytes = 3;            code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
           ptr += 3;  
4183            }            }
4184    
4185          /* 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
4186          if a digit follows ( then there will just be digits until ) because          syntax (?(<name>) or (?('name') */
         the syntax was checked in the first pass. */  
4187    
4188          else if ((digitab[ptr[1]] && ctype_digit) != 0)          else if (ptr[1] == '<')
4189            {            {
4190            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;  
             }  
4191            ptr++;            ptr++;
           code[1+LINK_SIZE] = OP_CREF;  
           PUT2(code, 2+LINK_SIZE, condref);  
           skipbytes = 3;  
4192            }            }
4193          /* For conditions that are assertions, we just fall through, having          else if (ptr[1] == '\'')
         set bravalue above. */  
         break;  
   
         case '=':                 /* Positive lookahead */  
         bravalue = OP_ASSERT;  
         ptr++;  
         break;  
   
         case '!':                 /* Negative lookahead */  
         bravalue = OP_ASSERT_NOT;  
         ptr++;  
         break;  
   
         case '<':                 /* Lookbehinds */  
         switch (*(++ptr))  
4194            {            {
4195            case '=':               /* Positive lookbehind */            terminator = '\'';
           bravalue = OP_ASSERTBACK;  
4196            ptr++;            ptr++;
4197            break;            }
4198            else
4199              {
4200              terminator = 0;
4201              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4202              }
4203    
4204            case '!':               /* Negative lookbehind */          /* We now expect to read a name; any thing else is an error */
4205            bravalue = OP_ASSERTBACK_NOT;  
4206            ptr++;          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
4207            break;            {
4208              ptr += 1;  /* To get the right offset */
4209              *errorcodeptr = ERR28;
4210              goto FAILED;
4211            }            }
         break;  
4212    
4213          case '>':                 /* One-time brackets */          /* Read the name, but also get it as a number if it's all digits */
         bravalue = OP_ONCE;  
         ptr++;  
         break;  
4214    
4215          case 'C':                 /* Callout - may be followed by digits; */          recno = 0;
4216          previous_callout = code;  /* Save for later completion */          name = ++ptr;
4217          after_manual_callout = 1; /* Skip one item before completing */          while ((cd->ctypes[*ptr] & ctype_word) != 0)
4218          *code++ = OP_CALLOUT;     /* Already checked that the terminating */            {
4219            {                       /* closing parenthesis is present. */            if (recno >= 0)
4220            int n = 0;              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4221            while ((digitab[*(++ptr)] & ctype_digit) != 0)                recno * 10 + *ptr - '0' : -1;
4222              n = n * 10 + *ptr - '0';            ptr++;
           if (n > 255)  
             {  
             *errorcodeptr = ERR38;  
             goto FAILED;  
             }  
           *code++ = n;  
           PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */  
           PUT(code, LINK_SIZE, 0);                    /* Default length */  
           code += 2 * LINK_SIZE;  
4223            }            }
4224          previous = NULL;          namelen = ptr - name;
         continue;  
4225    
4226          case 'P':                 /* Named subpattern handling */          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
         if (*(++ptr) == '<')      /* Definition */  
4227            {            {
4228            int i, namelen;            ptr--;      /* Error offset */
4229            uschar *slot = cd->name_table;            *errorcodeptr = ERR26;
4230            const uschar *name;     /* Don't amalgamate; some compilers */            goto FAILED;
4231            name = ++ptr;           /* grumble at autoincrement in declaration */            }
4232    
4233            while (*ptr++ != '>');          /* Do no further checking in the pre-compile phase. */
           namelen = ptr - name - 1;  
4234    
4235            for (i = 0; i < cd->names_found; i++)          if (lengthptr != NULL) break;
4236              {  
4237              int crc = memcmp(name, slot+2, namelen);          /* In the real compile we do the work of looking for the actual
4238              if (crc == 0)          reference. If the string started with "+" or "-" we require the rest to
4239                {          be digits, in which case recno will be set. */
4240                if (slot[2+namelen] == 0)  
4241            if (refsign > 0)
4242              {
4243              if (recno <= 0)
4244                {
4245                *errorcodeptr = ERR58;
4246                goto FAILED;
4247                }
4248              recno = (refsign == '-')?
4249                cd->bracount - recno + 1 : recno +cd->bracount;
4250              if (recno <= 0 || recno > cd->final_bracount)
4251                {
4252                *errorcodeptr = ERR15;
4253                goto FAILED;
4254                }
4255              PUT2(code, 2+LINK_SIZE, recno);
4256              break;
4257              }
4258    
4259            /* Otherwise (did not start with "+" or "-"), start by looking for the
4260            name. */
4261    
4262            slot = cd->name_table;
4263            for (i = 0; i < cd->names_found; i++)
4264              {
4265              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
4266              slot += cd->name_entry_size;
4267              }
4268    
4269            /* Found a previous named subpattern */
4270    
4271            if (i < cd->names_found)
4272              {
4273              recno = GET2(slot, 0);
4274              PUT2(code, 2+LINK_SIZE, recno);
4275              }
4276    
4277            /* Search the pattern for a forward reference */
4278    
4279            else if ((i = find_parens(ptr, cd->bracount, name, namelen,
4280                            (options & PCRE_EXTENDED) != 0)) > 0)
4281              {
4282              PUT2(code, 2+LINK_SIZE, i);
4283              }
4284    
4285            /* If terminator == 0 it means that the name followed directly after
4286            the opening parenthesis [e.g. (?(abc)...] and in this case there are
4287            some further alternatives to try. For the cases where terminator != 0
4288            [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
4289            now checked all the possibilities, so give an error. */
4290    
4291            else if (terminator != 0)
4292              {
4293              *errorcodeptr = ERR15;
4294              goto FAILED;
4295              }
4296    
4297            /* Check for (?(R) for recursion. Allow digits after R to specify a
4298            specific group number. */
4299    
4300            else if (*name == 'R')
4301              {
4302              recno = 0;
4303              for (i = 1; i < namelen; i++)
4304                {
4305                if ((digitab[name[i]] & ctype_digit) == 0)
4306                  {
4307                  *errorcodeptr = ERR15;
4308                  goto FAILED;
4309                  }
4310                recno = recno * 10 + name[i] - '0';
4311                }
4312              if (recno == 0) recno = RREF_ANY;
4313              code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
4314              PUT2(code, 2+LINK_SIZE, recno);
4315              }
4316    
4317            /* Similarly, check for the (?(DEFINE) "condition", which is always
4318            false. */
4319    
4320            else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)
4321              {
4322              code[1+LINK_SIZE] = OP_DEF;
4323              skipbytes = 1;
4324              }
4325    
4326            /* Check for the "name" actually being a subpattern number. We are
4327            in the second pass here, so final_bracount is set. */
4328    
4329            else if (recno > 0 && recno <= cd->final_bracount)
4330              {
4331              PUT2(code, 2+LINK_SIZE, recno);
4332              }
4333    
4334            /* Either an unidentified subpattern, or a reference to (?(0) */
4335    
4336            else
4337              {
4338              *errorcodeptr = (recno == 0)? ERR35: ERR15;
4339              goto FAILED;
4340              }
4341            break;
4342    
4343    
4344            /* ------------------------------------------------------------ */
4345            case '=':                 /* Positive lookahead */
4346            bravalue = OP_ASSERT;
4347            ptr++;
4348            break;
4349    
4350    
4351            /* ------------------------------------------------------------ */
4352            case '!':                 /* Negative lookahead */
4353            ptr++;
4354            if (*ptr == ')')          /* Optimize (?!) */
4355              {
4356              *code++ = OP_FAIL;
4357              previous = NULL;
4358              continue;
4359              }
4360            bravalue = OP_ASSERT_NOT;
4361            break;
4362    
4363    
4364            /* ------------------------------------------------------------ */
4365            case '<':                 /* Lookbehind or named define */
4366            switch (ptr[1])
4367              {
4368              case '=':               /* Positive lookbehind */
4369              bravalue = OP_ASSERTBACK;
4370              ptr += 2;
4371              break;
4372    
4373              case '!':               /* Negative lookbehind */
4374              bravalue = OP_ASSERTBACK_NOT;
4375              ptr += 2;
4376              break;
4377    
4378              default:                /* Could be name define, else bad */
4379              if ((cd->ctypes[ptr[1]] & ctype_word) != 0) goto DEFINE_NAME;
4380              ptr++;                  /* Correct offset for error */
4381              *errorcodeptr = ERR24;
4382              goto FAILED;
4383              }
4384            break;
4385    
4386    
4387            /* ------------------------------------------------------------ */
4388            case '>':                 /* One-time brackets */
4389            bravalue = OP_ONCE;
4390            ptr++;
4391            break;
4392    
4393    
4394            /* ------------------------------------------------------------ */
4395            case 'C':                 /* Callout - may be followed by digits; */
4396            previous_callout = code;  /* Save for later completion */
4397            after_manual_callout = 1; /* Skip one item before completing */
4398            *code++ = OP_CALLOUT;
4399              {
4400              int n = 0;
4401              while ((digitab[*(++ptr)] & ctype_digit) != 0)
4402                n = n * 10 + *ptr - '0';
4403              if (*ptr != ')')
4404                {
4405                *errorcodeptr = ERR39;
4406                goto FAILED;
4407                }
4408              if (n > 255)
4409                {
4410                *errorcodeptr = ERR38;
4411                goto FAILED;
4412                }
4413              *code++ = n;
4414              PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */