/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 79 by nigel, Sat Feb 24 21:40:52 2007 UTC revision 295 by ph10, Mon Dec 31 17:00:24 2007 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2007 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56    /* When DEBUG is defined, we need the pcre_printint() function, which is also
57    used by pcretest. DEBUG is not defined when building a production library. */
58    
59    #ifdef DEBUG
60    #include "pcre_printint.src"
61    #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
79    
80  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
81  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
82  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
83  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
84  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
85    so this number is very generous.
86    
87    The same workspace is used during the second, actual compile phase for
88    remembering forward references to groups so that they can be filled in at the
89    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
90    is 4 there is plenty of room. */
91    
92  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
93    
94    
95  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 63  are simple data values; negative values Line 97  are simple data values; negative values
97  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
98  is invalid. */  is invalid. */
99    
100  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */
101  static const short int escapes[] = {  static const short int escapes[] = {
102       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */
103       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */
104     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */
105       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */
106  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */
107  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */
108     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */
109       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */
110  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */
111       0,      0, -ESC_z                                            /* x - z */       0,      0, -ESC_z                                            /* x - z */
112  };  };
113    
114  #else         /* This is the "abnormal" table for EBCDIC systems */  #else           /* This is the "abnormal" table for EBCDIC systems */
115  static const short int escapes[] = {  static const short int escapes[] = {
116  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
117  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 121  static const short int escapes[] = {
121  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
122  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
123  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
124  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
125  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
126  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
127  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
128  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
129  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
130  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
131  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
132  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
133  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
134  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
135  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
136  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
137  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
138  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143  /* 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 220  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 256  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 270  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 304  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 331  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 342  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 355  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 388  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 401  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 422  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 442  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 560  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 587  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 619  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 698  read_repeat_counts(const uschar *p, int Line 897  read_repeat_counts(const uschar *p, int
897  int min = 0;  int min = 0;
898  int max = -1;  int max = -1;
899    
900    /* Read the minimum value and do a paranoid check: a negative value indicates
901    an integer overflow. */
902    
903  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';
904    if (min < 0 || min > 65535)
905      {
906      *errorcodeptr = ERR5;
907      return p;
908      }
909    
910    /* Read the maximum value if there is one, and again do a paranoid on its size.
911    Also, max must not be less than min. */
912    
913  if (*p == '}') max = min; else  if (*p == '}') max = min; else
914    {    {
# Line 706  if (*p == '}') max = min; else Line 916  if (*p == '}') max = min; else
916      {      {
917      max = 0;      max = 0;
918      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';
919        if (max < 0 || max > 65535)
920          {
921          *errorcodeptr = ERR5;
922          return p;
923          }
924      if (max < min)      if (max < min)
925        {        {
926        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 929  if (*p == '}') max = min; else
929      }      }
930    }    }
931    
932  /* Do paranoid checks, then fill in the required variables, and pass back the  /* Fill in the required variables, and pass back the pointer to the terminating
933  pointer to the terminating '}'. */  '}'. */
934    
935  if (min > 65535 || max > 65535)  *minp = min;
936    *errorcodeptr = ERR5;  *maxp = max;
937  else  return p;
938    }
939    
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    *minp = min;    int term;
972    *maxp = max;  
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  return p;  
1049    return -1;
1050  }  }
1051    
1052    
# Line 778  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 823  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 865  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 884  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 910  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 917  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 998  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 1031  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 1072  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 1103  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 1132  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.  backward and negative forward assertions when its final argument is TRUE. If we
1512    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1513    bracket whose current branch will already have been scanned.
1514    
1515  Arguments:  Arguments:
1516    code        points to start of search    code        points to start of search
# Line 1149  static BOOL Line 1524  static BOOL
1524  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1525  {  {
1526  register int c;  register int c;
1527  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);
1528       code < endcode;       code < endcode;
1529       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1530    {    {
# Line 1157  for (code = first_significant_code(code Line 1532  for (code = first_significant_code(code
1532    
1533    c = *code;    c = *code;
1534    
1535    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1536      first_significant_code() with a TRUE final argument. */
1537    
1538      if (c == OP_ASSERT)
1539        {
1540        do code += GET(code, 1); while (*code == OP_ALT);
1541        c = *code;
1542        continue;
1543        }
1544    
1545      /* Groups with zero repeats can of course be empty; skip them. */
1546    
1547      if (c == OP_BRAZERO || c == OP_BRAMINZERO)
1548        {
1549        code += _pcre_OP_lengths[c];
1550        do code += GET(code, 1); while (*code == OP_ALT);
1551        c = *code;
1552        continue;
1553        }
1554    
1555      /* For other groups, scan the branches. */
1556    
1557      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1558      {      {
1559      BOOL empty_branch;      BOOL empty_branch;
1560      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1173  for (code = first_significant_code(code Line 1570  for (code = first_significant_code(code
1570        }        }
1571      while (*code == OP_ALT);      while (*code == OP_ALT);
1572      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
     code += 1 + LINK_SIZE;  
1573      c = *code;      c = *code;
1574        continue;
1575      }      }
1576    
1577    else switch (c)    /* Handle the other opcodes */
1578    
1579      switch (c)
1580      {      {
1581      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1582        cannot be represented just by a bit map. This includes negated single
1583        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1584        actual length is stored in the compiled code, so we must update "code"
1585        here. */
1586    
1587  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1588      case OP_XCLASS:      case OP_XCLASS:
1589      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1590      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1591  #endif  #endif
1592    
# Line 1233  for (code = first_significant_code(code Line 1636  for (code = first_significant_code(code
1636      case OP_NOT:      case OP_NOT:
1637      case OP_PLUS:      case OP_PLUS:
1638      case OP_MINPLUS:      case OP_MINPLUS:
1639        case OP_POSPLUS:
1640      case OP_EXACT:      case OP_EXACT:
1641      case OP_NOTPLUS:      case OP_NOTPLUS:
1642      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1643        case OP_NOTPOSPLUS:
1644      case OP_NOTEXACT:      case OP_NOTEXACT:
1645      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1646      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1647        case OP_TYPEPOSPLUS:
1648      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1649      return FALSE;      return FALSE;
1650    
1651        /* These are going to continue, as they may be empty, but we have to
1652        fudge the length for the \p and \P cases. */
1653    
1654        case OP_TYPESTAR:
1655        case OP_TYPEMINSTAR:
1656        case OP_TYPEPOSSTAR:
1657        case OP_TYPEQUERY:
1658        case OP_TYPEMINQUERY:
1659        case OP_TYPEPOSQUERY:
1660        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1661        break;
1662    
1663        /* Same for these */
1664    
1665        case OP_TYPEUPTO:
1666        case OP_TYPEMINUPTO:
1667        case OP_TYPEPOSUPTO:
1668        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1669        break;
1670    
1671      /* End of branch */      /* End of branch */
1672    
1673      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 1676  for (code = first_significant_code(code
1676      case OP_ALT:      case OP_ALT:
1677      return TRUE;      return TRUE;
1678    
1679      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1680      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1681    
1682  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1683      case OP_STAR:      case OP_STAR:
1684      case OP_MINSTAR:      case OP_MINSTAR:
1685        case OP_POSSTAR:
1686      case OP_QUERY:      case OP_QUERY:
1687      case OP_MINQUERY:      case OP_MINQUERY:
1688        case OP_POSQUERY:
1689      case OP_UPTO:      case OP_UPTO:
1690      case OP_MINUPTO:      case OP_MINUPTO:
1691        case OP_POSUPTO:
1692      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1693      break;      break;
1694  #endif  #endif
# Line 1308  return TRUE; Line 1737  return TRUE;
1737  *************************************************/  *************************************************/
1738    
1739  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1740  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
1741  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1742  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1743    
1744    Originally, this function only recognized a sequence of letters between the
1745    terminators, but it seems that Perl recognizes any sequence of characters,
1746    though of course unknown POSIX names are subsequently rejected. Perl gives an
1747    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1748    didn't consider this to be a POSIX class. Likewise for [:1234:].
1749    
1750    The problem in trying to be exactly like Perl is in the handling of escapes. We
1751    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1752    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1753    below handles the special case of \], but does not try to do any other escape
1754    processing. This makes it different from Perl for cases such as [:l\ower:]
1755    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1756    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1757    I think.
1758    
1759  Argument:  Arguments:
1760    ptr      pointer to the initial [    ptr      pointer to the initial [
1761    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1762    
1763  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1764  */  */
1765    
1766  static BOOL  static BOOL
1767  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1768  {  {
1769  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1770  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1771  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1772    {    {
1773    *endptr = ptr;    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1774    return TRUE;      {
1775    }      if (*ptr == ']') return FALSE;
1776        if (*ptr == terminator && ptr[1] == ']')
1777          {
1778          *endptr = ptr;
1779          return TRUE;
1780          }
1781        }
1782      }
1783  return FALSE;  return FALSE;
1784  }  }
1785    
# Line 1355  Returns:     a value representing the na Line 1803  Returns:     a value representing the na
1803  static int  static int
1804  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1805  {  {
1806    const char *pn = posix_names;
1807  register int yield = 0;  register int yield = 0;
1808  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1809    {    {
1810    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1811      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1812      pn += posix_name_lengths[yield] + 1;
1813    yield++;    yield++;
1814    }    }
1815  return -1;  return -1;
# Line 1377  earlier groups that are outside the curr Line 1827  earlier groups that are outside the curr
1827  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
1828  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
1829  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
1830  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,
1831  partially compiled regex must be temporarily terminated with OP_END.  the partially compiled regex must be temporarily terminated with OP_END.
1832    
1833    This function has been extended with the possibility of forward references for
1834    recursions and subroutine calls. It must also check the list of such references
1835    for the group we are dealing with. If it finds that one of the recursions in
1836    the current group is on this list, it adjusts the offset in the list, not the
1837    value in the reference (which is a group number).
1838    
1839  Arguments:  Arguments:
1840    group      points to the start of the group    group      points to the start of the group
1841    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1842    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1843    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1844      save_hwm   the hwm forward reference pointer at the start of the group
1845    
1846  Returns:     nothing  Returns:     nothing
1847  */  */
1848    
1849  static void  static void
1850  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1851      uschar *save_hwm)
1852  {  {
1853  uschar *ptr = group;  uschar *ptr = group;
1854    
1855  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1856    {    {
1857    int offset = GET(ptr, 1);    int offset;
1858    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1859    
1860      /* See if this recursion is on the forward reference list. If so, adjust the
1861      reference. */
1862    
1863      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1864        {
1865        offset = GET(hc, 0);
1866        if (cd->start_code + offset == ptr + 1)
1867          {
1868          PUT(hc, 0, offset + adjust);
1869          break;
1870          }
1871        }
1872    
1873      /* Otherwise, adjust the recursion offset if it's after the start of this
1874      group. */
1875    
1876      if (hc >= cd->hwm)
1877        {
1878        offset = GET(ptr, 1);
1879        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1880        }
1881    
1882    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1883    }    }
1884  }  }
# Line 1475  Yield:        TRUE when range returned; Line 1957  Yield:        TRUE when range returned;
1957  */  */
1958    
1959  static BOOL  static BOOL
1960  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
1961      unsigned int *odptr)
1962  {  {
1963  int c, chartype, othercase, next;  unsigned int c, othercase, next;
1964    
1965  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
1966    {    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
1967    
1968  if (c > d) return FALSE;  if (c > d) return FALSE;
1969    
# Line 1492  next = othercase + 1; Line 1972  next = othercase + 1;
1972    
1973  for (++c; c <= d; c++)  for (++c; c <= d; c++)
1974    {    {
1975    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (_pcre_ucp_othercase(c) != next) break;
         othercase != next)  
     break;  
1976    next++;    next++;
1977    }    }
1978    
# Line 1506  return TRUE; Line 1984  return TRUE;
1984  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
1985    
1986    
1987    
1988  /*************************************************  /*************************************************
1989  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
1990  *************************************************/  *************************************************/
1991    
1992  /* 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
1993  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
1994  bits.  sense to automatically possessify the repeated item.
1995    
1996  Arguments:  Arguments:
1997    optionsptr     pointer to the option bits    op_code       the repeated op code
1998    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
1999    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
2000    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
2001    errorcodeptr   points to error code variable    ptr           next character in pattern
2002    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
2003    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.  
2004    
2005  Returns:         TRUE on success  Returns:        TRUE if possessifying is wanted
                  FALSE, with *errorcodeptr set non-zero on error  
2006  */  */
2007    
2008  static BOOL  static BOOL
2009  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2010    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const uschar *ptr, int options, compile_data *cd)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
2011  {  {
2012  int repeat_type, op_type;  int next;
2013  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  
2014  int bravalue = 0;  /* Skip whitespace and comments in extended mode */
2015    
2016    if ((options & PCRE_EXTENDED) != 0)
2017      {
2018      for (;;)
2019        {
2020        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2021        if (*ptr == '#')
2022          {
2023          while (*(++ptr) != 0)
2024            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2025          }
2026        else break;
2027        }
2028      }
2029    
2030    /* If the next item is one that we can handle, get its value. A non-negative
2031    value is a character, a negative value is an escape value. */
2032    
2033    if (*ptr == '\\')
2034      {
2035      int temperrorcode = 0;
2036      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2037      if (temperrorcode != 0) return FALSE;
2038      ptr++;    /* Point after the escape sequence */
2039      }
2040    
2041    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2042      {
2043    #ifdef SUPPORT_UTF8
2044      if (utf8) { GETCHARINC(next, ptr); } else
2045    #endif
2046      next = *ptr++;
2047      }
2048    
2049    else return FALSE;
2050    
2051    /* Skip whitespace and comments in extended mode */
2052    
2053    if ((options & PCRE_EXTENDED) != 0)
2054      {
2055      for (;;)
2056        {
2057        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2058        if (*ptr == '#')
2059          {
2060          while (*(++ptr) != 0)
2061            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2062          }
2063        else break;
2064        }
2065      }
2066    
2067    /* If the next thing is itself optional, we have to give up. */
2068    
2069    if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
2070      return FALSE;
2071    
2072    /* Now compare the next item with the previous opcode. If the previous is a
2073    positive single character match, "item" either contains the character or, if
2074    "item" is greater than 127 in utf8 mode, the character's bytes are in
2075    utf8_char. */
2076    
2077    
2078    /* Handle cases when the next item is a character. */
2079    
2080    if (next >= 0) switch(op_code)
2081      {
2082      case OP_CHAR:
2083    #ifdef SUPPORT_UTF8
2084      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2085    #endif
2086      return item != next;
2087    
2088      /* For CHARNC (caseless character) we must check the other case. If we have
2089      Unicode property support, we can use it to test the other case of
2090      high-valued characters. */
2091    
2092      case OP_CHARNC:
2093    #ifdef SUPPORT_UTF8
2094      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2095    #endif
2096      if (item == next) return FALSE;
2097    #ifdef SUPPORT_UTF8
2098      if (utf8)
2099        {
2100        unsigned int othercase;
2101        if (next < 128) othercase = cd->fcc[next]; else
2102    #ifdef SUPPORT_UCP
2103        othercase = _pcre_ucp_othercase((unsigned int)next);
2104    #else
2105        othercase = NOTACHAR;
2106    #endif
2107        return (unsigned int)item != othercase;
2108        }
2109      else
2110    #endif  /* SUPPORT_UTF8 */
2111      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2112    
2113      /* For OP_NOT, "item" must be a single-byte character. */
2114    
2115      case OP_NOT:
2116      if (next < 0) return FALSE;  /* Not a character */
2117      if (item == next) return TRUE;
2118      if ((options & PCRE_CASELESS) == 0) return FALSE;
2119    #ifdef SUPPORT_UTF8
2120      if (utf8)
2121        {
2122        unsigned int othercase;
2123        if (next < 128) othercase = cd->fcc[next]; else
2124    #ifdef SUPPORT_UCP
2125        othercase = _pcre_ucp_othercase(next);
2126    #else
2127        othercase = NOTACHAR;
2128    #endif
2129        return (unsigned int)item == othercase;
2130        }
2131      else
2132    #endif  /* SUPPORT_UTF8 */
2133      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2134    
2135      case OP_DIGIT:
2136      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2137    
2138      case OP_NOT_DIGIT:
2139      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2140    
2141      case OP_WHITESPACE:
2142      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2143    
2144      case OP_NOT_WHITESPACE:
2145      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2146    
2147      case OP_WORDCHAR:
2148      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2149    
2150      case OP_NOT_WORDCHAR:
2151      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2152    
2153      case OP_HSPACE:
2154      case OP_NOT_HSPACE:
2155      switch(next)
2156        {
2157        case 0x09:
2158        case 0x20:
2159        case 0xa0:
2160        case 0x1680:
2161        case 0x180e:
2162        case 0x2000:
2163        case 0x2001:
2164        case 0x2002:
2165        case 0x2003:
2166        case 0x2004:
2167        case 0x2005:
2168        case 0x2006:
2169        case 0x2007:
2170        case 0x2008:
2171        case 0x2009:
2172        case 0x200A:
2173        case 0x202f:
2174        case 0x205f:
2175        case 0x3000:
2176        return op_code != OP_HSPACE;
2177        default:
2178        return op_code == OP_HSPACE;
2179        }
2180    
2181      case OP_VSPACE:
2182      case OP_NOT_VSPACE:
2183      switch(next)
2184        {
2185        case 0x0a:
2186        case 0x0b:
2187        case 0x0c:
2188        case 0x0d:
2189        case 0x85:
2190        case 0x2028:
2191        case 0x2029:
2192        return op_code != OP_VSPACE;
2193        default:
2194        return op_code == OP_VSPACE;
2195        }
2196    
2197      default:
2198      return FALSE;
2199      }
2200    
2201    
2202    /* Handle the case when the next item is \d, \s, etc. */
2203    
2204    switch(op_code)
2205      {
2206      case OP_CHAR:
2207      case OP_CHARNC:
2208    #ifdef SUPPORT_UTF8
2209      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2210    #endif
2211      switch(-next)
2212        {
2213        case ESC_d:
2214        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2215    
2216        case ESC_D:
2217        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2218    
2219        case ESC_s:
2220        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2221    
2222        case ESC_S:
2223        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2224    
2225        case ESC_w:
2226        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2227    
2228        case ESC_W:
2229        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2230    
2231        case ESC_h:
2232        case ESC_H:
2233        switch(item)
2234          {
2235          case 0x09:
2236          case 0x20:
2237          case 0xa0:
2238          case 0x1680:
2239          case 0x180e:
2240          case 0x2000:
2241          case 0x2001:
2242          case 0x2002:
2243          case 0x2003:
2244          case 0x2004:
2245          case 0x2005:
2246          case 0x2006:
2247          case 0x2007:
2248          case 0x2008:
2249          case 0x2009:
2250          case 0x200A:
2251          case 0x202f:
2252          case 0x205f:
2253          case 0x3000:
2254          return -next != ESC_h;
2255          default:
2256          return -next == ESC_h;
2257          }
2258    
2259        case ESC_v:
2260        case ESC_V:
2261        switch(item)
2262          {
2263          case 0x0a:
2264          case 0x0b:
2265          case 0x0c:
2266          case 0x0d:
2267          case 0x85:
2268          case 0x2028:
2269          case 0x2029:
2270          return -next != ESC_v;
2271          default:
2272          return -next == ESC_v;
2273          }
2274    
2275        default:
2276        return FALSE;
2277        }
2278    
2279      case OP_DIGIT:
2280      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2281             next == -ESC_h || next == -ESC_v;
2282    
2283      case OP_NOT_DIGIT:
2284      return next == -ESC_d;
2285    
2286      case OP_WHITESPACE:
2287      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2288    
2289      case OP_NOT_WHITESPACE:
2290      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2291    
2292      case OP_HSPACE:
2293      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2294    
2295      case OP_NOT_HSPACE:
2296      return next == -ESC_h;
2297    
2298      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2299      case OP_VSPACE:
2300      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2301    
2302      case OP_NOT_VSPACE:
2303      return next == -ESC_v;
2304    
2305      case OP_WORDCHAR:
2306      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2307    
2308      case OP_NOT_WORDCHAR:
2309      return next == -ESC_w || next == -ESC_d;
2310    
2311      default:
2312      return FALSE;
2313      }
2314    
2315    /* Control does not reach here */
2316    }
2317    
2318    
2319    
2320    /*************************************************
2321    *           Compile one branch                   *
2322    *************************************************/
2323    
2324    /* Scan the pattern, compiling it into the a vector. If the options are
2325    changed during the branch, the pointer is used to change the external options
2326    bits. This function is used during the pre-compile phase when we are trying
2327    to find out the amount of memory needed, as well as during the real compile
2328    phase. The value of lengthptr distinguishes the two phases.
2329    
2330    Arguments:
2331      optionsptr     pointer to the option bits
2332      codeptr        points to the pointer to the current code point
2333      ptrptr         points to the current pattern pointer
2334      errorcodeptr   points to error code variable
2335      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2336      reqbyteptr     set to the last literal character required, else < 0
2337      bcptr          points to current branch chain
2338      cd             contains pointers to tables etc.
2339      lengthptr      NULL during the real compile phase
2340                     points to length accumulator during pre-compile phase
2341    
2342    Returns:         TRUE on success
2343                     FALSE, with *errorcodeptr set non-zero on error
2344    */
2345    
2346    static BOOL
2347    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2348      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2349      compile_data *cd, int *lengthptr)
2350    {
2351    int repeat_type, op_type;
2352    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2353    int bravalue = 0;
2354  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
2355  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2356  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2357  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2358  int options = *optionsptr;  int options = *optionsptr;
2359  int after_manual_callout = 0;  int after_manual_callout = 0;
2360    int length_prevgroup = 0;
2361  register int c;  register int c;
2362  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2363    uschar *last_code = code;
2364    uschar *orig_code = code;
2365  uschar *tempcode;  uschar *tempcode;
2366  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2367  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1553  const uschar *ptr = *ptrptr; Line 2369  const uschar *ptr = *ptrptr;
2369  const uschar *tempptr;  const uschar *tempptr;
2370  uschar *previous = NULL;  uschar *previous = NULL;
2371  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2372    uschar *save_hwm = NULL;
2373  uschar classbits[32];  uschar classbits[32];
2374    
2375  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1562  uschar *class_utf8data; Line 2379  uschar *class_utf8data;
2379  uschar utf8_char[6];  uschar utf8_char[6];
2380  #else  #else
2381  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2382    uschar *utf8_char = NULL;
2383    #endif
2384    
2385    #ifdef DEBUG
2386    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2387  #endif  #endif
2388    
2389  /* Set up the default and non-default settings for greediness */  /* Set up the default and non-default settings for greediness */
# Line 1593  req_caseopt = ((options & PCRE_CASELESS) Line 2415  req_caseopt = ((options & PCRE_CASELESS)
2415  for (;; ptr++)  for (;; ptr++)
2416    {    {
2417    BOOL negate_class;    BOOL negate_class;
2418      BOOL should_flip_negation;
2419    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2420    BOOL is_quantifier;    BOOL is_quantifier;
2421      BOOL is_recurse;
2422      BOOL reset_bracount;
2423    int class_charcount;    int class_charcount;
2424    int class_lastchar;    int class_lastchar;
2425    int newoptions;    int newoptions;
2426    int recno;    int recno;
2427      int refsign;
2428    int skipbytes;    int skipbytes;
2429    int subreqbyte;    int subreqbyte;
2430    int subfirstbyte;    int subfirstbyte;
2431      int terminator;
2432    int mclength;    int mclength;
2433    uschar mcbuffer[8];    uschar mcbuffer[8];
2434    
2435    /* Next byte in the pattern */    /* Get next byte in the pattern */
2436    
2437    c = *ptr;    c = *ptr;
2438    
2439      /* If we are in the pre-compile phase, accumulate the length used for the
2440      previous cycle of this loop. */
2441    
2442      if (lengthptr != NULL)
2443        {
2444    #ifdef DEBUG
2445        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2446    #endif
2447        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2448          {
2449          *errorcodeptr = ERR52;
2450          goto FAILED;
2451          }
2452    
2453        /* There is at least one situation where code goes backwards: this is the
2454        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2455        the class is simply eliminated. However, it is created first, so we have to
2456        allow memory for it. Therefore, don't ever reduce the length at this point.
2457        */
2458    
2459        if (code < last_code) code = last_code;
2460    
2461        /* Paranoid check for integer overflow */
2462    
2463        if (OFLOW_MAX - *lengthptr < code - last_code)
2464          {
2465          *errorcodeptr = ERR20;
2466          goto FAILED;
2467          }
2468    
2469        *lengthptr += code - last_code;
2470        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2471    
2472        /* If "previous" is set and it is not at the start of the work space, move
2473        it back to there, in order to avoid filling up the work space. Otherwise,
2474        if "previous" is NULL, reset the current code pointer to the start. */
2475    
2476        if (previous != NULL)
2477          {
2478          if (previous > orig_code)
2479            {
2480            memmove(orig_code, previous, code - previous);
2481            code -= previous - orig_code;
2482            previous = orig_code;
2483            }
2484          }
2485        else code = orig_code;
2486    
2487        /* Remember where this code item starts so we can pick up the length
2488        next time round. */
2489    
2490        last_code = code;
2491        }
2492    
2493      /* In the real compile phase, just check the workspace used by the forward
2494      reference list. */
2495    
2496      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2497        {
2498        *errorcodeptr = ERR52;
2499        goto FAILED;
2500        }
2501    
2502    /* 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 */
2503    
2504    if (inescq && c != 0)    if (inescq && c != 0)
# Line 1623  for (;; ptr++) Line 2513  for (;; ptr++)
2513        {        {
2514        if (previous_callout != NULL)        if (previous_callout != NULL)
2515          {          {
2516          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2517              complete_callout(previous_callout, ptr, cd);
2518          previous_callout = NULL;          previous_callout = NULL;
2519          }          }
2520        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1644  for (;; ptr++) Line 2535  for (;; ptr++)
2535    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2536         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2537      {      {
2538      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2539          complete_callout(previous_callout, ptr, cd);
2540      previous_callout = NULL;      previous_callout = NULL;
2541      }      }
2542    
# Line 1655  for (;; ptr++) Line 2547  for (;; ptr++)
2547      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2548      if (c == '#')      if (c == '#')
2549        {        {
2550        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2551        on the Macintosh. */          {
2552        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2553        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2554          if (*ptr != 0) continue;
2555    
2556          /* Else fall through to handle end of string */
2557          c = 0;
2558        }        }
2559      }      }
2560    
# Line 1672  for (;; ptr++) Line 2568  for (;; ptr++)
2568    
2569    switch(c)    switch(c)
2570      {      {
2571      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2572        case 0:                        /* The branch terminates at string end */
2573      case 0:      case '|':                      /* or | or ) */
     case '|':  
2574      case ')':      case ')':
2575      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2576      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2577      *codeptr = code;      *codeptr = code;
2578      *ptrptr = ptr;      *ptrptr = ptr;
2579        if (lengthptr != NULL)
2580          {
2581          if (OFLOW_MAX - *lengthptr < code - last_code)
2582            {
2583            *errorcodeptr = ERR20;
2584            goto FAILED;
2585            }
2586          *lengthptr += code - last_code;   /* To include callout length */
2587          DPRINTF((">> end branch\n"));
2588          }
2589      return TRUE;      return TRUE;
2590    
2591    
2592        /* ===================================================================*/
2593      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2594      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2595    
# Line 1711  for (;; ptr++) Line 2618  for (;; ptr++)
2618      *code++ = OP_ANY;      *code++ = OP_ANY;
2619      break;      break;
2620    
2621      /* Character classes. If the included characters are all < 255 in value, we  
2622      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2623      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
2624      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
2625      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2626        map as usual, then invert it at the end. However, we use a different opcode
2627        so that data characters > 255 can be handled correctly.
2628    
2629      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2630      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 1730  for (;; ptr++) Line 2639  for (;; ptr++)
2639      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
2640    
2641      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2642          check_posix_syntax(ptr, &tempptr, cd))          check_posix_syntax(ptr, &tempptr))
2643        {        {
2644        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;
2645        goto FAILED;        goto FAILED;
2646        }        }
2647    
2648      /* 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,
2649        if the first few characters (either before or after ^) are \Q\E or \E we
2650        skip them too. This makes for compatibility with Perl. */
2651    
2652      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2653        for (;;)
2654        {        {
       negate_class = TRUE;  
2655        c = *(++ptr);        c = *(++ptr);
2656          if (c == '\\')
2657            {
2658            if (ptr[1] == 'E') ptr++;
2659              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2660                else break;
2661            }
2662          else if (!negate_class && c == '^')
2663            negate_class = TRUE;
2664          else break;
2665        }        }
2666      else  
2667        {      /* If a class contains a negative special such as \S, we need to flip the
2668        negate_class = FALSE;      negation flag at the end, so that support for characters > 255 works
2669        }      correctly (they are all included in the class). */
2670    
2671        should_flip_negation = FALSE;
2672    
2673      /* 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
2674      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
2675      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2676    
2677      class_charcount = 0;      class_charcount = 0;
2678      class_lastchar = -1;      class_lastchar = -1;
2679    
2680        /* Initialize the 32-char bit map to all zeros. We build the map in a
2681        temporary bit of memory, in case the class contains only 1 character (less
2682        than 256), because in that case the compiled code doesn't use the bit map.
2683        */
2684    
2685        memset(classbits, 0, 32 * sizeof(uschar));
2686    
2687  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2688      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2689      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2690  #endif  #endif
2691    
     /* 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));  
   
2692      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2693      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
2694      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. */  
2695    
2696      do      if (c != 0) do
2697        {        {
2698          const uschar *oldptr;
2699    
2700  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2701        if (utf8 && c > 127)        if (utf8 && c > 127)
2702          {                           /* Braces are required because the */          {                           /* Braces are required because the */
# Line 1786  for (;; ptr++) Line 2708  for (;; ptr++)
2708    
2709        if (inescq)        if (inescq)
2710          {          {
2711          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2712            {            {
2713            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2714            ptr++;            ptr++;                            /* Skip the 'E' */
2715            continue;            continue;                         /* Carry on with next */
2716            }            }
2717          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2718          }          }
2719    
2720        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1803  for (;; ptr++) Line 2725  for (;; ptr++)
2725    
2726        if (c == '[' &&        if (c == '[' &&
2727            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2728            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr))
2729          {          {
2730          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2731          int posix_class, i;          int posix_class, taboffset, tabopt;
2732          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2733            uschar pbits[32];
2734    
2735          if (ptr[1] != ':')          if (ptr[1] != ':')
2736            {            {
# Line 1819  for (;; ptr++) Line 2742  for (;; ptr++)
2742          if (*ptr == '^')          if (*ptr == '^')
2743            {            {
2744            local_negate = TRUE;            local_negate = TRUE;
2745              should_flip_negation = TRUE;  /* Note negative special */
2746            ptr++;            ptr++;
2747            }            }
2748    
# Line 1836  for (;; ptr++) Line 2760  for (;; ptr++)
2760          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
2761            posix_class = 0;            posix_class = 0;
2762    
2763          /* 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
2764          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
2765          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
2766          white space chars afterwards. */          result into the bit map that is being built. */
2767    
2768          posix_class *= 3;          posix_class *= 3;
2769          for (i = 0; i < 3; i++)  
2770            /* Copy in the first table (always present) */
2771    
2772            memcpy(pbits, cbits + posix_class_maps[posix_class],
2773              32 * sizeof(uschar));
2774    
2775            /* If there is a second table, add or remove it as required. */
2776    
2777            taboffset = posix_class_maps[posix_class + 1];
2778            tabopt = posix_class_maps[posix_class + 2];
2779    
2780            if (taboffset >= 0)
2781            {            {
2782            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
2783            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;  
             }  
2784            else            else
2785              {              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;  
             }  
2786            }            }
2787    
2788            /* Not see if we need to remove any special characters. An option
2789            value of 1 removes vertical space and 2 removes underscore. */
2790    
2791            if (tabopt < 0) tabopt = -tabopt;
2792            if (tabopt == 1) pbits[1] &= ~0x3c;
2793              else if (tabopt == 2) pbits[11] &= 0x7f;
2794    
2795            /* Add the POSIX table or its complement into the main table that is
2796            being built and we are done. */
2797    
2798            if (local_negate)
2799              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
2800            else
2801              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
2802    
2803          ptr = tempptr + 1;          ptr = tempptr + 1;
2804          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
2805          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
2806          }          }
2807    
2808        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2809        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
2810        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.
2811        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2812        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  
2813        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2814    
2815        if (c == '\\')        if (c == '\\')
2816          {          {
2817          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2818            if (*errorcodeptr != 0) goto FAILED;
2819    
2820          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2821          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 */
2822            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2823          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2824            {            {
2825            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1890  for (;; ptr++) Line 2829  for (;; ptr++)
2829            else inescq = TRUE;            else inescq = TRUE;
2830            continue;            continue;
2831            }            }
2832            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2833    
2834          if (c < 0)          if (c < 0)
2835            {            {
2836            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2837            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2838            switch (-c)  
2839              /* Save time by not doing this in the pre-compile phase. */
2840    
2841              if (lengthptr == NULL) switch (-c)
2842              {              {
2843              case ESC_d:              case ESC_d:
2844              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
2845              continue;              continue;
2846    
2847              case ESC_D:              case ESC_D:
2848                should_flip_negation = TRUE;
2849              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2850              continue;              continue;
2851    
# Line 1910  for (;; ptr++) Line 2854  for (;; ptr++)
2854              continue;              continue;
2855    
2856              case ESC_W:              case ESC_W:
2857                should_flip_negation = TRUE;
2858              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2859              continue;              continue;
2860    
# Line 1919  for (;; ptr++) Line 2864  for (;; ptr++)
2864              continue;              continue;
2865    
2866              case ESC_S:              case ESC_S:
2867                should_flip_negation = TRUE;
2868              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2869              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2870              continue;              continue;
2871    
2872  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
2873              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
2874              case ESC_P:              }
2875    
2876              /* In the pre-compile phase, just do the recognition. */
2877    
2878              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2879                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
2880    
2881              /* We need to deal with \H, \h, \V, and \v in both phases because
2882              they use extra memory. */
2883    
2884              if (-c == ESC_h)
2885                {
2886                SETBIT(classbits, 0x09); /* VT */
2887                SETBIT(classbits, 0x20); /* SPACE */
2888                SETBIT(classbits, 0xa0); /* NSBP */
2889    #ifdef SUPPORT_UTF8
2890                if (utf8)
2891                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
2892                class_utf8 = TRUE;                class_utf8 = TRUE;
2893                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_SINGLE;
2894                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2895                *class_utf8data++ = property;                *class_utf8data++ = XCL_SINGLE;
2896                class_charcount -= 2;   /* Not a < 256 character */                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2897                  *class_utf8data++ = XCL_RANGE;
2898                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2899                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2900                  *class_utf8data++ = XCL_SINGLE;
2901                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2902                  *class_utf8data++ = XCL_SINGLE;
2903                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2904                  *class_utf8data++ = XCL_SINGLE;
2905                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2906                }                }
             continue;  
2907  #endif  #endif
2908                continue;
2909                }
2910    
2911              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_H)
2912              strict mode. By default, for compatibility with Perl, they are              {
2913              treated as literals. */              for (c = 0; c < 32; c++)
2914                  {
2915                  int x = 0xff;
2916                  switch (c)
2917                    {
2918                    case 0x09/8: x ^= 1 << (0x09%8); break;
2919                    case 0x20/8: x ^= 1 << (0x20%8); break;
2920                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2921                    default: break;
2922                    }
2923                  classbits[c] |= x;
2924                  }
2925    
2926              default:  #ifdef SUPPORT_UTF8
2927              if ((options & PCRE_EXTRA) != 0)              if (utf8)
2928                {                {
2929                *errorcodeptr = ERR7;                class_utf8 = TRUE;
2930                goto FAILED;                *class_utf8data++ = XCL_RANGE;
2931                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2932                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
2933                  *class_utf8data++ = XCL_RANGE;
2934                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
2935                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
2936                  *class_utf8data++ = XCL_RANGE;
2937                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
2938                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
2939                  *class_utf8data++ = XCL_RANGE;
2940                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
2941                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
2942                  *class_utf8data++ = XCL_RANGE;
2943                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
2944                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
2945                  *class_utf8data++ = XCL_RANGE;
2946                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
2947                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
2948                  *class_utf8data++ = XCL_RANGE;
2949                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
2950                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2951                }                }
2952              c = *ptr;              /* The final character */  #endif
2953              class_charcount -= 2;  /* Undo the default count from above */              continue;
2954              }              }
           }  
   
         /* Fall through if we have a single character (c >= 0). This may be  
         > 256 in UTF-8 mode. */  
2955    
2956          }   /* End of backslash handling */            if (-c == ESC_v)
2957                {
2958                SETBIT(classbits, 0x0a); /* LF */
2959                SETBIT(classbits, 0x0b); /* VT */
2960                SETBIT(classbits, 0x0c); /* FF */
2961                SETBIT(classbits, 0x0d); /* CR */
2962                SETBIT(classbits, 0x85); /* NEL */
2963    #ifdef SUPPORT_UTF8
2964                if (utf8)
2965                  {
2966                  class_utf8 = TRUE;
2967                  *class_utf8data++ = XCL_RANGE;
2968                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
2969                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2970                  }
2971    #endif
2972                continue;
2973                }
2974    
2975        /* A single character may be followed by '-' to form a range. However,            if (-c == ESC_V)
2976        Perl does not permit ']' to be the end of the range. A '-' character              {
2977        here is treated as a literal. */              for (c = 0; c < 32; c++)
2978                  {
2979                  int x = 0xff;
2980                  switch (c)
2981                    {
2982                    case 0x0a/8: x ^= 1 << (0x0a%8);
2983                                 x ^= 1 << (0x0b%8);
2984                                 x ^= 1 << (0x0c%8);
2985                                 x ^= 1 << (0x0d%8);
2986                                 break;
2987                    case 0x85/8: x ^= 1 << (0x85%8); break;
2988                    default: break;
2989                    }
2990                  classbits[c] |= x;
2991                  }
2992    
2993        if (ptr[1] == '-' && ptr[2] != ']')  #ifdef SUPPORT_UTF8
2994                if (utf8)
2995                  {
2996                  class_utf8 = TRUE;
2997                  *class_utf8data++ = XCL_RANGE;
2998                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2999                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3000                  *class_utf8data++ = XCL_RANGE;
3001                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3002                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3003                  }
3004    #endif
3005                continue;
3006                }
3007    
3008              /* We need to deal with \P and \p in both phases. */
3009    
3010    #ifdef SUPPORT_UCP
3011              if (-c == ESC_p || -c == ESC_P)
3012                {
3013                BOOL negated;
3014                int pdata;
3015                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3016                if (ptype < 0) goto FAILED;
3017                class_utf8 = TRUE;
3018                *class_utf8data++ = ((-c == ESC_p) != negated)?
3019                  XCL_PROP : XCL_NOTPROP;
3020                *class_utf8data++ = ptype;
3021                *class_utf8data++ = pdata;
3022                class_charcount -= 2;   /* Not a < 256 character */
3023                continue;
3024                }
3025    #endif
3026              /* Unrecognized escapes are faulted if PCRE is running in its
3027              strict mode. By default, for compatibility with Perl, they are
3028              treated as literals. */
3029    
3030              if ((options & PCRE_EXTRA) != 0)
3031                {
3032                *errorcodeptr = ERR7;
3033                goto FAILED;
3034                }
3035    
3036              class_charcount -= 2;  /* Undo the default count from above */
3037              c = *ptr;              /* Get the final character and fall through */
3038              }
3039    
3040            /* Fall through if we have a single character (c >= 0). This may be
3041            greater than 256 in UTF-8 mode. */
3042    
3043            }   /* End of backslash handling */
3044    
3045          /* A single character may be followed by '-' to form a range. However,
3046          Perl does not permit ']' to be the end of the range. A '-' character
3047          at the end is treated as a literal. Perl ignores orphaned \E sequences
3048          entirely. The code for handling \Q and \E is messy. */
3049    
3050          CHECK_RANGE:
3051          while (ptr[1] == '\\' && ptr[2] == 'E')
3052            {
3053            inescq = FALSE;
3054            ptr += 2;
3055            }
3056    
3057          oldptr = ptr;
3058    
3059          /* Remember \r or \n */
3060    
3061          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3062    
3063          /* Check for range */
3064    
3065          if (!inescq && ptr[1] == '-')
3066          {          {
3067          int d;          int d;
3068          ptr += 2;          ptr += 2;
3069            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
3070    
3071            /* If we hit \Q (not followed by \E) at this point, go into escaped
3072            mode. */
3073    
3074            while (*ptr == '\\' && ptr[1] == 'Q')
3075              {
3076              ptr += 2;
3077              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
3078              inescq = TRUE;
3079              break;
3080              }
3081    
3082            if (*ptr == 0 || (!inescq && *ptr == ']'))
3083              {
3084              ptr = oldptr;
3085              goto LONE_SINGLE_CHARACTER;
3086              }
3087    
3088  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3089          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3098  for (;; ptr++)
3098          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
3099          in such circumstances. */          in such circumstances. */
3100    
3101          if (d == '\\')          if (!inescq && d == '\\')
3102            {            {
3103            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3104            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3105    
3106            /* \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
3107            was literal */            special means the '-' was literal */
3108    
3109            if (d < 0)            if (d < 0)
3110              {              {
3111              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = '\b';
3112              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = 'X';
3113                else if (d == -ESC_R) d = 'R'; else
3114                {                {
3115                ptr = oldptr - 2;                ptr = oldptr;
3116                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3117                }                }
3118              }              }
3119            }            }
3120    
3121          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3122          the pre-pass. Optimize one-character ranges */          one-character ranges */
3123    
3124            if (d < c)
3125              {
3126              *errorcodeptr = ERR8;
3127              goto FAILED;
3128              }
3129    
3130          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3131    
3132            /* Remember \r or \n */
3133    
3134            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3135    
3136          /* 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
3137          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3138          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2022  for (;; ptr++) Line 3150  for (;; ptr++)
3150  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3151            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3152              {              {
3153              int occ, ocd;              unsigned int occ, ocd;
3154              int cc = c;              unsigned int cc = c;
3155              int origd = d;              unsigned int origd = d;
3156              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3157                {                {
3158                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3159                      ocd <= (unsigned int)d)
3160                    continue;                          /* Skip embedded ranges */
3161    
3162                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3163                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3164                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3165                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3166                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3167                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3168                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3169                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3170                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3171                  d = ocd;                  d = ocd;
3172                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3214  for (;; ptr++)
3214          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
3215          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3216    
3217          for (; c <= d; c++)          class_charcount += d - c + 1;
3218            class_lastchar = d;
3219    
3220            /* We can save a bit of time by skipping this in the pre-compile. */
3221    
3222            if (lengthptr == NULL) for (; c <= d; c++)
3223            {            {
3224            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3225            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3227  for (;; ptr++)
3227              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3228              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3229              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3230            }            }
3231    
3232          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3250  for (;; ptr++)
3250  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3251          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3252            {            {
3253            int chartype;            unsigned int othercase;
3254            int othercase;            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3255              {              {
3256              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3257              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3276  for (;; ptr++)
3276          }          }
3277        }        }
3278    
3279      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3280      loop. This "while" is the end of the "do" above. */  
3281        while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
3282    
3283        if (c == 0)                          /* Missing terminating ']' */
3284          {
3285          *errorcodeptr = ERR6;
3286          goto FAILED;
3287          }
3288    
3289    
3290    /* This code has been disabled because it would mean that \s counts as
3291    an explicit \r or \n reference, and that's not really what is wanted. Now
3292    we set the flag only if there is a literal "\r" or "\n" in the class. */
3293    
3294    #if 0
3295        /* Remember whether \r or \n are in this class */
3296    
3297        if (negate_class)
3298          {
3299          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3300          }
3301        else
3302          {
3303          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3304          }
3305    #endif
3306    
     while ((c = *(++ptr)) != ']' || inescq);  
3307    
3308      /* 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
3309      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
3310      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
3311      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3312      single-bytes only. This is an historical hangover. Maybe one day we can  
3313      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3314        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3315        operate on single-bytes only. This is an historical hangover. Maybe one day
3316        we can tidy these opcodes to handle multi-byte characters.
3317    
3318      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
3319      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2163  for (;; ptr++) Line 3323  for (;; ptr++)
3323      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3324    
3325  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3326      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3327            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3328  #else  #else
3329      if (class_charcount == 1)      if (class_charcount == 1)
3330  #endif  #endif
# Line 2209  for (;; ptr++) Line 3367  for (;; ptr++)
3367      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3368    
3369      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3370      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3371      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3372        the class, so any that were explicitly given as well can be ignored. If
3373        (when there are explicit characters > 255 that must be listed) there are no
3374        characters < 256, we can omit the bitmap in the actual compiled code. */
3375    
3376  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3377      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3378        {        {
3379        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3380        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3381        code += LINK_SIZE;        code += LINK_SIZE;
3382        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3383    
3384        /* 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;
3385        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3386    
3387        if (class_charcount > 0)        if (class_charcount > 0)
3388          {          {
3389          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3390            memmove(code + 32, code, class_utf8data - code);
3391          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3392          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;  
3393          }          }
3394          else code = class_utf8data;
3395    
3396        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3397    
# Line 2246  for (;; ptr++) Line 3400  for (;; ptr++)
3400        }        }
3401  #endif  #endif
3402    
3403      /* 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
3404      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
3405      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
3406      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3407    
3408        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3409      if (negate_class)      if (negate_class)
3410        {        {
3411        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3412        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3413        }        }
3414      else      else
3415        {        {
       *code++ = OP_CLASS;  
3416        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3417        }        }
3418      code += 32;      code += 32;
3419      break;      break;
3420    
3421    
3422        /* ===================================================================*/
3423      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3424      has been tested above. */      has been tested above. */
3425    
# Line 2331  for (;; ptr++) Line 3487  for (;; ptr++)
3487        }        }
3488      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3489    
     /* 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;  
       }  
   
3490      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3491      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
3492      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
# Line 2378  for (;; ptr++) Line 3520  for (;; ptr++)
3520          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3521          }          }
3522    
3523          /* If the repetition is unlimited, it pays to see if the next thing on
3524          the line is something that cannot possibly match this character. If so,
3525          automatically possessifying this item gains some performance in the case
3526          where the match fails. */
3527    
3528          if (!possessive_quantifier &&
3529              repeat_max < 0 &&
3530              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3531                options, cd))
3532            {
3533            repeat_type = 0;    /* Force greedy */
3534            possessive_quantifier = TRUE;
3535            }
3536    
3537        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3538        }        }
3539    
3540      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3541      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-
3542      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3543      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3544        currently used only for single-byte chars. */
3545    
3546      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3547        {        {
3548        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3549        c = previous[1];        c = previous[1];
3550          if (!possessive_quantifier &&
3551              repeat_max < 0 &&
3552              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3553            {
3554            repeat_type = 0;    /* Force greedy */
3555            possessive_quantifier = TRUE;
3556            }
3557        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3558        }        }
3559    
# Line 2403  for (;; ptr++) Line 3567  for (;; ptr++)
3567      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3568        {        {
3569        uschar *oldcode;        uschar *oldcode;
3570        int prop_type;        int prop_type, prop_value;
3571        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3572        c = *previous;        c = *previous;
3573    
3574          if (!possessive_quantifier &&
3575              repeat_max < 0 &&
3576              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3577            {
3578            repeat_type = 0;    /* Force greedy */
3579            possessive_quantifier = TRUE;
3580            }
3581    
3582        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3583        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3584          previous[1] : -1;          {
3585            prop_type = previous[1];
3586            prop_value = previous[2];
3587            }
3588          else prop_type = prop_value = -1;
3589    
3590        oldcode = code;        oldcode = code;
3591        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2422  for (;; ptr++) Line 3598  for (;; ptr++)
3598        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3599        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3600    
3601        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3602    
3603        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3604    
# Line 2443  for (;; ptr++) Line 3619  for (;; ptr++)
3619          }          }
3620    
3621        /* 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
3622        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3623        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
3624        one less than the maximum. */        one less than the maximum. */
3625    
# Line 2470  for (;; ptr++) Line 3646  for (;; ptr++)
3646    
3647          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3648          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
3649          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3650          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3651          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3652    
# Line 2486  for (;; ptr++) Line 3662  for (;; ptr++)
3662  #endif  #endif
3663              {              {
3664              *code++ = c;              *code++ = c;
3665              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3666                  {
3667                  *code++ = prop_type;
3668                  *code++ = prop_value;
3669                  }
3670              }              }
3671            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3672            }            }
3673    
3674          /* 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
3675          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3676            UPTO is just for 1 instance, we can use QUERY instead. */
3677    
3678          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3679            {            {
# Line 2505  for (;; ptr++) Line 3686  for (;; ptr++)
3686            else            else
3687  #endif  #endif
3688            *code++ = c;            *code++ = c;
3689            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3690                {
3691                *code++ = prop_type;
3692                *code++ = prop_value;
3693                }
3694            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3695            *code++ = OP_UPTO + repeat_type;  
3696            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3697                {
3698                *code++ = OP_QUERY + repeat_type;
3699                }
3700              else
3701                {
3702                *code++ = OP_UPTO + repeat_type;
3703                PUT2INC(code, 0, repeat_max);
3704                }
3705            }            }
3706          }          }
3707    
# Line 2524  for (;; ptr++) Line 3717  for (;; ptr++)
3717  #endif  #endif
3718        *code++ = c;        *code++ = c;
3719    
3720        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3721        defines the required property. */        define the required property. */
3722    
3723  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3724        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3725            {
3726            *code++ = prop_type;
3727            *code++ = prop_value;
3728            }
3729  #endif  #endif
3730        }        }
3731    
# Line 2551  for (;; ptr++) Line 3748  for (;; ptr++)
3748        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3749        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3750    
3751        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3752    
3753        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3754          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2571  for (;; ptr++) Line 3768  for (;; ptr++)
3768      /* 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
3769      cases. */      cases. */
3770    
3771      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3772               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3773        {        {
3774        register int i;        register int i;
3775        int ketoffset = 0;        int ketoffset = 0;
3776        int len = code - previous;        int len = code - previous;
3777        uschar *bralink = NULL;        uschar *bralink = NULL;
3778    
3779          /* Repeating a DEFINE group is pointless */
3780    
3781          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3782            {
3783            *errorcodeptr = ERR55;
3784            goto FAILED;
3785            }
3786    
3787        /* 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
3788        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
3789        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 2613  for (;; ptr++) Line 3818  for (;; ptr++)
3818          /* 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
3819          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
3820          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
3821          any internal group, because the offset is from the start of the whole          any internal or forward referenced group, because the offset is from
3822          regex. Temporarily terminate the pattern while doing this. */          the start of the whole regex. Temporarily terminate the pattern while
3823            doing this. */
3824    
3825          if (repeat_max <= 1)          if (repeat_max <= 1)
3826            {            {
3827            *code = OP_END;            *code = OP_END;
3828            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3829            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3830            code++;            code++;
3831            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2637  for (;; ptr++) Line 3843  for (;; ptr++)
3843            {            {
3844            int offset;            int offset;
3845            *code = OP_END;            *code = OP_END;
3846            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3847            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3848            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3849            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 3863  for (;; ptr++)
3863        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3864        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3865        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
3866        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3867          forward reference subroutine calls in the group, there will be entries on
3868          the workspace list; replicate these with an appropriate increment. */
3869    
3870        else        else
3871          {          {
3872          if (repeat_min > 1)          if (repeat_min > 1)
3873            {            {
3874            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3875            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
3876              potential integer overflow. */
3877    
3878              if (lengthptr != NULL)
3879                {
3880                int delta = (repeat_min - 1)*length_prevgroup;
3881                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3882                                                                (double)INT_MAX ||
3883                    OFLOW_MAX - *lengthptr < delta)
3884                  {
3885                  *errorcodeptr = ERR20;
3886                  goto FAILED;
3887                  }
3888                *lengthptr += delta;
3889                }
3890    
3891              /* This is compiling for real */
3892    
3893              else
3894              {              {
3895              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3896              code += len;              for (i = 1; i < repeat_min; i++)
3897                  {
3898                  uschar *hc;
3899                  uschar *this_hwm = cd->hwm;
3900                  memcpy(code, previous, len);
3901                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3902                    {
3903                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3904                    cd->hwm += LINK_SIZE;
3905                    }
3906                  save_hwm = this_hwm;
3907                  code += len;
3908                  }
3909              }              }
3910            }            }
3911    
3912          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3913          }          }
3914    
# Line 2677  for (;; ptr++) Line 3916  for (;; ptr++)
3916        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3917        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,
3918        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
3919        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
3920          replicate entries on the forward reference list. */
3921    
3922        if (repeat_max >= 0)        if (repeat_max >= 0)
3923          {          {
3924          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
3925            just adjust the length as if we had. For each repetition we must add 1
3926            to the length for BRAZERO and for all but the last repetition we must
3927            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3928            paranoid checks to avoid integer overflow. */
3929    
3930            if (lengthptr != NULL && repeat_max > 0)
3931              {
3932              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3933                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3934              if ((double)repeat_max *
3935                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3936                      > (double)INT_MAX ||
3937                  OFLOW_MAX - *lengthptr < delta)
3938                {
3939                *errorcodeptr = ERR20;
3940                goto FAILED;
3941                }
3942              *lengthptr += delta;
3943              }
3944    
3945            /* This is compiling for real */
3946    
3947            else for (i = repeat_max - 1; i >= 0; i--)
3948            {            {
3949              uschar *hc;
3950              uschar *this_hwm = cd->hwm;
3951    
3952            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
3953    
3954            /* All but the final copy start a new nesting, maintaining the            /* All but the final copy start a new nesting, maintaining the
# Line 2698  for (;; ptr++) Line 3964  for (;; ptr++)
3964              }              }
3965    
3966            memcpy(code, previous, len);            memcpy(code, previous, len);
3967              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3968                {
3969                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
3970                cd->hwm += LINK_SIZE;
3971                }
3972              save_hwm = this_hwm;
3973            code += len;            code += len;
3974            }            }
3975    
# Line 2720  for (;; ptr++) Line 3992  for (;; ptr++)
3992        /* 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
3993        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
3994        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
3995        correct offset was computed above. */        correct offset was computed above.
3996    
3997          Then, when we are doing the actual compile phase, check to see whether
3998          this group is a non-atomic one that could match an empty string. If so,
3999          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
4000          that runtime checking can be done. [This check is also applied to
4001          atomic groups at runtime, but in a different way.] */
4002    
4003        else code[-ketoffset] = OP_KETRMAX + repeat_type;        else
4004            {
4005            uschar *ketcode = code - ketoffset;
4006            uschar *bracode = ketcode - GET(ketcode, 1);
4007            *ketcode = OP_KETRMAX + repeat_type;
4008            if (lengthptr == NULL && *bracode != OP_ONCE)
4009              {
4010              uschar *scode = bracode;
4011              do
4012                {
4013                if (could_be_empty_branch(scode, ketcode, utf8))
4014                  {
4015                  *bracode += OP_SBRA - OP_BRA;
4016                  break;
4017                  }
4018                scode += GET(scode, 1);
4019                }
4020              while (*scode == OP_ALT);
4021              }
4022            }
4023        }        }
4024    
4025      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2733  for (;; ptr++) Line 4030  for (;; ptr++)
4030        goto FAILED;        goto FAILED;
4031        }        }
4032    
4033      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
4034      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
4035      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
4036      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.
4037      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
4038        but the special opcodes can optimize it a bit. The repeated item starts at
4039        tempcode, not at previous, which might be the first part of a string whose
4040        (former) last char we repeated.
4041    
4042        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4043        an 'upto' may follow. We skip over an 'exact' item, and then test the
4044        length of what remains before proceeding. */
4045    
4046      if (possessive_quantifier)      if (possessive_quantifier)
4047        {        {
4048        int len = code - tempcode;        int len;
4049        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4050        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
4051        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode] +
4052        tempcode[0] = OP_ONCE;            ((*tempcode == OP_TYPEEXACT &&
4053        *code++ = OP_KET;               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4054        PUTINC(code, 0, len);        len = code - tempcode;
4055        PUT(tempcode, 1, len);        if (len > 0) switch (*tempcode)
4056            {
4057            case OP_STAR:  *tempcode = OP_POSSTAR; break;
4058            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
4059            case OP_QUERY: *tempcode = OP_POSQUERY; break;
4060            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4061    
4062            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4063            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4064            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4065            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4066    
4067            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4068            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4069            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4070            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4071    
4072            default:
4073            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4074            code += 1 + LINK_SIZE;
4075            len += 1 + LINK_SIZE;
4076            tempcode[0] = OP_ONCE;
4077            *code++ = OP_KET;
4078            PUTINC(code, 0, len);
4079            PUT(tempcode, 1, len);
4080            break;
4081            }
4082        }        }
4083    
4084      /* In all case we no longer have a previous item. We also set the      /* In all case we no longer have a previous item. We also set the
# Line 2761  for (;; ptr++) Line 4091  for (;; ptr++)
4091      break;      break;
4092    
4093    
4094      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
4095      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4096      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
4097      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.  */  
4098    
4099      case '(':      case '(':
4100      newoptions = options;      newoptions = options;
4101      skipbytes = 0;      skipbytes = 0;
4102        bravalue = OP_CBRA;
4103        save_hwm = cd->hwm;
4104        reset_bracount = FALSE;
4105    
4106        /* First deal with various "verbs" that can be introduced by '*'. */
4107    
4108        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4109          {
4110          int i, namelen;
4111          const char *vn = verbnames;
4112          const uschar *name = ++ptr;
4113          previous = NULL;
4114          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4115          if (*ptr == ':')
4116            {
4117            *errorcodeptr = ERR59;   /* Not supported */
4118            goto FAILED;
4119            }
4120          if (*ptr != ')')
4121            {
4122            *errorcodeptr = ERR60;
4123            goto FAILED;
4124            }
4125          namelen = ptr - name;
4126          for (i = 0; i < verbcount; i++)
4127            {
4128            if (namelen == verbs[i].len &&
4129                strncmp((char *)name, vn, namelen) == 0)
4130              {
4131              *code = verbs[i].op;
4132              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4133              break;
4134              }
4135            vn += verbs[i].len + 1;
4136            }
4137          if (i < verbcount) continue;
4138          *errorcodeptr = ERR60;
4139          goto FAILED;
4140          }
4141    
4142      if (*(++ptr) == '?')      /* Deal with the extended parentheses; all are introduced by '?', and the
4143        appearance of any of them means that this is not a capturing group. */
4144    
4145        else if (*ptr == '?')
4146        {        {
4147        int set, unset;        int i, set, unset, namelen;
4148        int *optset;        int *optset;
4149          const uschar *name;
4150          uschar *slot;
4151    
4152        switch (*(++ptr))        switch (*(++ptr))
4153          {          {
4154          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
4155          ptr++;          ptr++;
4156          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
4157            if (*ptr == 0)
4158              {
4159              *errorcodeptr = ERR18;
4160              goto FAILED;
4161              }
4162          continue;          continue;
4163    
4164          case ':':                 /* Non-extracting bracket */  
4165            /* ------------------------------------------------------------ */
4166            case '|':                 /* Reset capture count for each branch */
4167            reset_bracount = TRUE;
4168            /* Fall through */
4169    
4170            /* ------------------------------------------------------------ */
4171            case ':':                 /* Non-capturing bracket */
4172          bravalue = OP_BRA;          bravalue = OP_BRA;
4173          ptr++;          ptr++;
4174          break;          break;
4175    
4176    
4177            /* ------------------------------------------------------------ */
4178          case '(':          case '(':
4179          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4180    
4181          /* Condition to test for recursion */          /* A condition can be an assertion, a number (referring to a numbered
4182            group), a name (referring to a named group), or 'R', referring to
4183            recursion. R<digits> and R&name are also permitted for recursion tests.
4184    
4185            There are several syntaxes for testing a named group: (?(name)) is used
4186            by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
4187    
4188            There are two unfortunate ambiguities, caused by history. (a) 'R' can
4189            be the recursive thing or the name 'R' (and similarly for 'R' followed
4190            by digits), and (b) a number could be a name that consists of digits.
4191            In both cases, we look for a name first; if not found, we try the other
4192            cases. */
4193    
4194            /* For conditions that are assertions, check the syntax, and then exit
4195            the switch. This will take control down to where bracketed groups,
4196            including assertions, are processed. */
4197    
4198          if (ptr[1] == 'R')          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
4199              break;
4200    
4201            /* Most other conditions use OP_CREF (a couple change to OP_RREF
4202            below), and all need to skip 3 bytes at the start of the group. */
4203    
4204            code[1+LINK_SIZE] = OP_CREF;
4205            skipbytes = 3;
4206            refsign = -1;
4207    
4208            /* Check for a test for recursion in a named group. */
4209    
4210            if (ptr[1] == 'R' && ptr[2] == '&')
4211            {            {
4212            code[1+LINK_SIZE] = OP_CREF;            terminator = -1;
4213            PUT2(code, 2+LINK_SIZE, CREF_RECURSE);            ptr += 2;
4214            skipbytes = 3;            code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
           ptr += 3;  
4215            }            }
4216    
4217          /* 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
4218          if a digit follows ( then there will just be digits until ) because          syntax (?(<name>) or (?('name') */
         the syntax was checked in the first pass. */  
4219    
4220          else if ((digitab[ptr[1]] && ctype_digit) != 0)          else if (ptr[1] == '<')
4221            {            {
4222            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;  
             }  
4223            ptr++;            ptr++;
           code[1+LINK_SIZE] = OP_CREF;  
           PUT2(code, 2+LINK_SIZE, condref);  
           skipbytes = 3;  
4224            }            }
4225          /* 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))  
4226            {            {
4227            case '=':               /* Positive lookbehind */            terminator = '\'';
           bravalue = OP_ASSERTBACK;  
4228            ptr++;            ptr++;
4229            break;            }
4230            else
4231              {
4232              terminator = 0;
4233              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4234              }
4235    
4236            case '!':               /* Negative lookbehind */          /* We now expect to read a name; any thing else is an error */
4237            bravalue = OP_ASSERTBACK_NOT;  
4238            ptr++;          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
4239            break;            {
4240              ptr += 1;  /* To get the right offset */
4241              *errorcodeptr = ERR28;
4242              goto FAILED;
4243            }            }
         break;  
4244    
4245          case '>':                 /* One-time brackets */          /* Read the name, but also get it as a number if it's all digits */
         bravalue = OP_ONCE;  
         ptr++;  
         break;  
4246    
4247          case 'C':                 /* Callout - may be followed by digits; */          recno = 0;
4248          previous_callout = code;  /* Save for later completion */          name = ++ptr;
4249          after_manual_callout = 1; /* Skip one item before completing */          while ((cd->ctypes[*ptr] & ctype_word) != 0)
4250          *code++ = OP_CALLOUT;     /* Already checked that the terminating */            {
4251            {                       /* closing parenthesis is present. */            if (recno >= 0)
4252            int n = 0;              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4253            while ((digitab[*(++ptr)] & ctype_digit) != 0)                recno * 10 + *ptr - '0' : -1;
4254              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;  
4255            }            }
4256          previous = NULL;          namelen = ptr - name;
         continue;  
4257    
4258          case 'P':                 /* Named subpattern handling */          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
         if (*(++ptr) == '<')      /* Definition */  
4259            {            {
4260            int i, namelen;            ptr--;      /* Error offset */
4261            uschar *slot = cd->name_table;            *errorcodeptr = ERR26;
4262            const uschar *name;     /* Don't amalgamate; some compilers */            goto FAILED;
4263            name = ++ptr;           /* grumble at autoincrement in declaration */            }
4264    
4265            while (*ptr++ != '>');          /* Do no further checking in the pre-compile phase. */
           namelen = ptr - name - 1;  
4266    
4267            for (i = 0; i < cd->names_found; i++)          if (lengthptr != NULL) break;
             {  
             int crc = memcmp(name, slot+2, namelen);  
             if (crc == 0)  
               {  
               if (slot[2+namelen] == 0)  
                 {  
                 *errorcodeptr = ERR43;  
                 goto FAILED;  
                 }  
               crc = -1;             /* Current name is substring */  
               }  
             if (crc < 0)  
               {  
               memmove(slot + cd->name_entry_size, slot,  
                 (cd->names_found - i) * cd->name_entry_size);  
               break;  
               }  
             slot += cd->name_entry_size;  
             }  
4268    
4269            PUT2(slot, 0, *brackets + 1);          /* In the real compile we do the work of looking for the actual
4270            memcpy(slot + 2, name, namelen);          reference. If the string started with "+" or "-" we require the rest to
4271            slot[2+namelen] = 0;          be digits, in which case recno will be set. */
           cd->names_found++;  
           goto NUMBERED_GROUP;  
           }  
4272    
4273          if (*ptr == '=' || *ptr == '>')  /* Reference or recursion */          if (refsign > 0)
4274            {            {
4275            int i, namelen;            if (recno <= 0)
           int type = *ptr++;  
           const uschar *name = ptr;  
           uschar *slot = cd->name_table;  
   
           while (*ptr != ')') ptr++;  
           namelen = ptr - name;  
   
           for (i = 0; i < cd->names_found; i++)  
4276              {              {
4277              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              *errorcodeptr = ERR58;
4278              slot += cd->name_entry_size;              goto FAILED;
4279              }              }
4280            if (i >= cd->names_found)            recno = (refsign == '-')?
4281                cd->bracount - recno + 1 : recno +cd->bracount;
4282              if (recno <= 0 || recno > cd->final_bracount)
4283              {              {
4284              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
4285              goto FAILED;              goto FAILED;
4286              }              }
4287              PUT2(code, 2+LINK_SIZE, recno);
4288              break;
4289              }
4290    
4291            /* Otherwise (did not start with "+" or "-"), start by looking for the
4292            name. */
4293    
4294            slot = cd->name_table;
4295            for (i = 0; i < cd->names_found; i++)
4296              {
4297              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
4298              slot += cd->name_entry_size;
4299              }
4300    
4301            /* Found a previous named subpattern */
4302    
4303            if (i < cd->names_found)
4304              {
4305            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4306              PUT2(code, 2+LINK_SIZE, recno);
4307              }
4308    
4309            if (type == '>') goto HANDLE_RECURSION;  /* A few lines below */          /* Search the pattern for a forward reference */
4310    
4311            else if ((i = find_parens(ptr, cd->bracount, name, namelen,
4312                            (options & PCRE_EXTENDED) != 0)) > 0)
4313              {
4314              PUT2(code, 2+LINK_SIZE, i);
4315              }
4316    
4317            /* Back reference */          /* If terminator == 0 it means that the name followed directly after
4318            the opening parenthesis [e.g. (?(abc)...] and in this case there are
4319            some further alternatives to try. For the cases where terminator != 0
4320            [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
4321            now checked all the possibilities, so give an error. */
4322