/[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 286 by ph10, Mon Dec 17 14:46:11 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 1355  Returns: a value representing the na Line 1784  Returns: a value representing the na
1784  static int  static int
1785  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1786  {  {
1787    const char *pn = posix_names;
1788  register int yield = 0;  register int yield = 0;
1789  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1790    {    {
1791    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1792      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1793      pn += posix_name_lengths[yield] + 1;
1794    yield++;    yield++;
1795    }    }
1796  return -1;  return -1;
# Line 1377  earlier groups that are outside the curr Line 1808  earlier groups that are outside the curr
1808  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
1809  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
1810  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
1811  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,
1812  partially compiled regex must be temporarily terminated with OP_END.  the partially compiled regex must be temporarily terminated with OP_END.
1813    
1814    This function has been extended with the possibility of forward references for
1815    recursions and subroutine calls. It must also check the list of such references
1816    for the group we are dealing with. If it finds that one of the recursions in
1817    the current group is on this list, it adjusts the offset in the list, not the
1818    value in the reference (which is a group number).
1819    
1820  Arguments:  Arguments:
1821    group      points to the start of the group    group      points to the start of the group
1822    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1823    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1824    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1825      save_hwm   the hwm forward reference pointer at the start of the group
1826    
1827  Returns:     nothing  Returns:     nothing
1828  */  */
1829    
1830  static void  static void
1831  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1832      uschar *save_hwm)
1833  {  {
1834  uschar *ptr = group;  uschar *ptr = group;
1835    
1836  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1837    {    {
1838    int offset = GET(ptr, 1);    int offset;
1839    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1840    
1841      /* See if this recursion is on the forward reference list. If so, adjust the
1842      reference. */
1843    
1844      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1845        {
1846        offset = GET(hc, 0);
1847        if (cd->start_code + offset == ptr + 1)
1848          {
1849          PUT(hc, 0, offset + adjust);
1850          break;
1851          }
1852        }
1853    
1854      /* Otherwise, adjust the recursion offset if it's after the start of this
1855      group. */
1856    
1857      if (hc >= cd->hwm)
1858        {
1859        offset = GET(ptr, 1);
1860        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1861        }
1862    
1863    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1864    }    }
1865  }  }
# Line 1475  Yield: TRUE when range returned; Line 1938  Yield: TRUE when range returned;
1938  */  */
1939    
1940  static BOOL  static BOOL
1941  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
1942      unsigned int *odptr)
1943  {  {
1944  int c, chartype, othercase, next;  unsigned int c, othercase, next;
1945    
1946  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
1947    {    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
1948    
1949  if (c > d) return FALSE;  if (c > d) return FALSE;
1950    
# Line 1492  next = othercase + 1; Line 1953  next = othercase + 1;
1953    
1954  for (++c; c <= d; c++)  for (++c; c <= d; c++)
1955    {    {
1956    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (_pcre_ucp_othercase(c) != next) break;
         othercase != next)  
     break;  
1957    next++;    next++;
1958    }    }
1959    
# Line 1506  return TRUE; Line 1965  return TRUE;
1965  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
1966    
1967    
1968    
1969  /*************************************************  /*************************************************
1970  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
1971  *************************************************/  *************************************************/
1972    
1973  /* 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
1974  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
1975  bits.  sense to automatically possessify the repeated item.
1976    
1977  Arguments:  Arguments:
1978    optionsptr     pointer to the option bits    op_code       the repeated op code
1979    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
1980    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
1981    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
1982    errorcodeptr   points to error code variable    ptr           next character in pattern
1983    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
1984    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.  
1985    
1986  Returns:         TRUE on success  Returns:        TRUE if possessifying is wanted
                  FALSE, with *errorcodeptr set non-zero on error  
1987  */  */
1988    
1989  static BOOL  static BOOL
1990  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
1991    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const uschar *ptr, int options, compile_data *cd)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
1992  {  {
1993  int repeat_type, op_type;  int next;
1994  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  
1995  int bravalue = 0;  /* Skip whitespace and comments in extended mode */
1996  int greedy_default, greedy_non_default;  
1997  int firstbyte, reqbyte;  if ((options & PCRE_EXTENDED) != 0)
1998  int zeroreqbyte, zerofirstbyte;    {
1999      for (;;)
2000        {
2001        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2002        if (*ptr == '#')
2003          {
2004          while (*(++ptr) != 0)
2005            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2006          }
2007        else break;
2008        }
2009      }
2010    
2011    /* If the next item is one that we can handle, get its value. A non-negative
2012    value is a character, a negative value is an escape value. */
2013    
2014    if (*ptr == '\\')
2015      {
2016      int temperrorcode = 0;
2017      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2018      if (temperrorcode != 0) return FALSE;
2019      ptr++;    /* Point after the escape sequence */
2020      }
2021    
2022    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2023      {
2024    #ifdef SUPPORT_UTF8
2025      if (utf8) { GETCHARINC(next, ptr); } else
2026    #endif
2027      next = *ptr++;
2028      }
2029    
2030    else return FALSE;
2031    
2032    /* Skip whitespace and comments in extended mode */
2033    
2034    if ((options & PCRE_EXTENDED) != 0)
2035      {
2036      for (;;)
2037        {
2038        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2039        if (*ptr == '#')
2040          {
2041          while (*(++ptr) != 0)
2042            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2043          }
2044        else break;
2045        }
2046      }
2047    
2048    /* If the next thing is itself optional, we have to give up. */
2049    
2050    if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
2051      return FALSE;
2052    
2053    /* Now compare the next item with the previous opcode. If the previous is a
2054    positive single character match, "item" either contains the character or, if
2055    "item" is greater than 127 in utf8 mode, the character's bytes are in
2056    utf8_char. */
2057    
2058    
2059    /* Handle cases when the next item is a character. */
2060    
2061    if (next >= 0) switch(op_code)
2062      {
2063      case OP_CHAR:
2064    #ifdef SUPPORT_UTF8
2065      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2066    #endif
2067      return item != next;
2068    
2069      /* For CHARNC (caseless character) we must check the other case. If we have
2070      Unicode property support, we can use it to test the other case of
2071      high-valued characters. */
2072    
2073      case OP_CHARNC:
2074    #ifdef SUPPORT_UTF8
2075      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2076    #endif
2077      if (item == next) return FALSE;
2078    #ifdef SUPPORT_UTF8
2079      if (utf8)
2080        {
2081        unsigned int othercase;
2082        if (next < 128) othercase = cd->fcc[next]; else
2083    #ifdef SUPPORT_UCP
2084        othercase = _pcre_ucp_othercase((unsigned int)next);
2085    #else
2086        othercase = NOTACHAR;
2087    #endif
2088        return (unsigned int)item != othercase;
2089        }
2090      else
2091    #endif  /* SUPPORT_UTF8 */
2092      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2093    
2094      /* For OP_NOT, "item" must be a single-byte character. */
2095    
2096      case OP_NOT:
2097      if (next < 0) return FALSE;  /* Not a character */
2098      if (item == next) return TRUE;
2099      if ((options & PCRE_CASELESS) == 0) return FALSE;
2100    #ifdef SUPPORT_UTF8
2101      if (utf8)
2102        {
2103        unsigned int othercase;
2104        if (next < 128) othercase = cd->fcc[next]; else
2105    #ifdef SUPPORT_UCP
2106        othercase = _pcre_ucp_othercase(next);
2107    #else
2108        othercase = NOTACHAR;
2109    #endif
2110        return (unsigned int)item == othercase;
2111        }
2112      else
2113    #endif  /* SUPPORT_UTF8 */
2114      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2115    
2116      case OP_DIGIT:
2117      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2118    
2119      case OP_NOT_DIGIT:
2120      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2121    
2122      case OP_WHITESPACE:
2123      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2124    
2125      case OP_NOT_WHITESPACE:
2126      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2127    
2128      case OP_WORDCHAR:
2129      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2130    
2131      case OP_NOT_WORDCHAR:
2132      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2133    
2134      case OP_HSPACE:
2135      case OP_NOT_HSPACE:
2136      switch(next)
2137        {
2138        case 0x09:
2139        case 0x20:
2140        case 0xa0:
2141        case 0x1680:
2142        case 0x180e:
2143        case 0x2000:
2144        case 0x2001:
2145        case 0x2002:
2146        case 0x2003:
2147        case 0x2004:
2148        case 0x2005:
2149        case 0x2006:
2150        case 0x2007:
2151        case 0x2008:
2152        case 0x2009:
2153        case 0x200A:
2154        case 0x202f:
2155        case 0x205f:
2156        case 0x3000:
2157        return op_code != OP_HSPACE;
2158        default:
2159        return op_code == OP_HSPACE;
2160        }
2161    
2162      case OP_VSPACE:
2163      case OP_NOT_VSPACE:
2164      switch(next)
2165        {
2166        case 0x0a:
2167        case 0x0b:
2168        case 0x0c:
2169        case 0x0d:
2170        case 0x85:
2171        case 0x2028:
2172        case 0x2029:
2173        return op_code != OP_VSPACE;
2174        default:
2175        return op_code == OP_VSPACE;
2176        }
2177    
2178      default:
2179      return FALSE;
2180      }
2181    
2182    
2183    /* Handle the case when the next item is \d, \s, etc. */
2184    
2185    switch(op_code)
2186      {
2187      case OP_CHAR:
2188      case OP_CHARNC:
2189    #ifdef SUPPORT_UTF8
2190      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2191    #endif
2192      switch(-next)
2193        {
2194        case ESC_d:
2195        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2196    
2197        case ESC_D:
2198        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2199    
2200        case ESC_s:
2201        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2202    
2203        case ESC_S:
2204        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2205    
2206        case ESC_w:
2207        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2208    
2209        case ESC_W:
2210        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2211    
2212        case ESC_h:
2213        case ESC_H:
2214        switch(item)
2215          {
2216          case 0x09:
2217          case 0x20:
2218          case 0xa0:
2219          case 0x1680:
2220          case 0x180e:
2221          case 0x2000:
2222          case 0x2001:
2223          case 0x2002:
2224          case 0x2003:
2225          case 0x2004:
2226          case 0x2005:
2227          case 0x2006:
2228          case 0x2007:
2229          case 0x2008:
2230          case 0x2009:
2231          case 0x200A:
2232          case 0x202f:
2233          case 0x205f:
2234          case 0x3000:
2235          return -next != ESC_h;
2236          default:
2237          return -next == ESC_h;
2238          }
2239    
2240        case ESC_v:
2241        case ESC_V:
2242        switch(item)
2243          {
2244          case 0x0a:
2245          case 0x0b:
2246          case 0x0c:
2247          case 0x0d:
2248          case 0x85:
2249          case 0x2028:
2250          case 0x2029:
2251          return -next != ESC_v;
2252          default:
2253          return -next == ESC_v;
2254          }
2255    
2256        default:
2257        return FALSE;
2258        }
2259    
2260      case OP_DIGIT:
2261      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2262             next == -ESC_h || next == -ESC_v;
2263    
2264      case OP_NOT_DIGIT:
2265      return next == -ESC_d;
2266    
2267      case OP_WHITESPACE:
2268      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2269    
2270      case OP_NOT_WHITESPACE:
2271      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2272    
2273      case OP_HSPACE:
2274      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2275    
2276      case OP_NOT_HSPACE:
2277      return next == -ESC_h;
2278    
2279      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2280      case OP_VSPACE:
2281      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2282    
2283      case OP_NOT_VSPACE:
2284      return next == -ESC_v;
2285    
2286      case OP_WORDCHAR:
2287      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2288    
2289      case OP_NOT_WORDCHAR:
2290      return next == -ESC_w || next == -ESC_d;
2291    
2292      default:
2293      return FALSE;
2294      }
2295    
2296    /* Control does not reach here */
2297    }
2298    
2299    
2300    
2301    /*************************************************
2302    *           Compile one branch                   *
2303    *************************************************/
2304    
2305    /* Scan the pattern, compiling it into the a vector. If the options are
2306    changed during the branch, the pointer is used to change the external options
2307    bits. This function is used during the pre-compile phase when we are trying
2308    to find out the amount of memory needed, as well as during the real compile
2309    phase. The value of lengthptr distinguishes the two phases.
2310    
2311    Arguments:
2312      optionsptr     pointer to the option bits
2313      codeptr        points to the pointer to the current code point
2314      ptrptr         points to the current pattern pointer
2315      errorcodeptr   points to error code variable
2316      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2317      reqbyteptr     set to the last literal character required, else < 0
2318      bcptr          points to current branch chain
2319      cd             contains pointers to tables etc.
2320      lengthptr      NULL during the real compile phase
2321                     points to length accumulator during pre-compile phase
2322    
2323    Returns:         TRUE on success
2324                     FALSE, with *errorcodeptr set non-zero on error
2325    */
2326    
2327    static BOOL
2328    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2329      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2330      compile_data *cd, int *lengthptr)
2331    {
2332    int repeat_type, op_type;
2333    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2334    int bravalue = 0;
2335    int greedy_default, greedy_non_default;
2336    int firstbyte, reqbyte;
2337    int zeroreqbyte, zerofirstbyte;
2338  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2339  int options = *optionsptr;  int options = *optionsptr;
2340  int after_manual_callout = 0;  int after_manual_callout = 0;
2341    int length_prevgroup = 0;
2342  register int c;  register int c;
2343  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2344    uschar *last_code = code;
2345    uschar *orig_code = code;
2346  uschar *tempcode;  uschar *tempcode;
2347  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2348  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1553  const uschar *ptr = *ptrptr; Line 2350  const uschar *ptr = *ptrptr;
2350  const uschar *tempptr;  const uschar *tempptr;
2351  uschar *previous = NULL;  uschar *previous = NULL;
2352  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2353    uschar *save_hwm = NULL;
2354  uschar classbits[32];  uschar classbits[32];
2355    
2356  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1562  uschar *class_utf8data; Line 2360  uschar *class_utf8data;
2360  uschar utf8_char[6];  uschar utf8_char[6];
2361  #else  #else
2362  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2363    uschar *utf8_char = NULL;
2364    #endif
2365    
2366    #ifdef DEBUG
2367    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2368  #endif  #endif
2369    
2370  /* 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 2396  req_caseopt = ((options & PCRE_CASELESS)
2396  for (;; ptr++)  for (;; ptr++)
2397    {    {
2398    BOOL negate_class;    BOOL negate_class;
2399      BOOL should_flip_negation;
2400    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2401    BOOL is_quantifier;    BOOL is_quantifier;
2402      BOOL is_recurse;
2403      BOOL reset_bracount;
2404    int class_charcount;    int class_charcount;
2405    int class_lastchar;    int class_lastchar;
2406    int newoptions;    int newoptions;
2407    int recno;    int recno;
2408      int refsign;
2409    int skipbytes;    int skipbytes;
2410    int subreqbyte;    int subreqbyte;
2411    int subfirstbyte;    int subfirstbyte;
2412      int terminator;
2413    int mclength;    int mclength;
2414    uschar mcbuffer[8];    uschar mcbuffer[8];
2415    
2416    /* Next byte in the pattern */    /* Get next byte in the pattern */
2417    
2418    c = *ptr;    c = *ptr;
2419    
2420      /* If we are in the pre-compile phase, accumulate the length used for the
2421      previous cycle of this loop. */
2422    
2423      if (lengthptr != NULL)
2424        {
2425    #ifdef DEBUG
2426        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2427    #endif
2428        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2429          {
2430          *errorcodeptr = ERR52;
2431          goto FAILED;
2432          }
2433    
2434        /* There is at least one situation where code goes backwards: this is the
2435        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2436        the class is simply eliminated. However, it is created first, so we have to
2437        allow memory for it. Therefore, don't ever reduce the length at this point.
2438        */
2439    
2440        if (code < last_code) code = last_code;
2441    
2442        /* Paranoid check for integer overflow */
2443    
2444        if (OFLOW_MAX - *lengthptr < code - last_code)
2445          {
2446          *errorcodeptr = ERR20;
2447          goto FAILED;
2448          }
2449    
2450        *lengthptr += code - last_code;
2451        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2452    
2453        /* If "previous" is set and it is not at the start of the work space, move
2454        it back to there, in order to avoid filling up the work space. Otherwise,
2455        if "previous" is NULL, reset the current code pointer to the start. */
2456    
2457        if (previous != NULL)
2458          {
2459          if (previous > orig_code)
2460            {
2461            memmove(orig_code, previous, code - previous);
2462            code -= previous - orig_code;
2463            previous = orig_code;
2464            }
2465          }
2466        else code = orig_code;
2467    
2468        /* Remember where this code item starts so we can pick up the length
2469        next time round. */
2470    
2471        last_code = code;
2472        }
2473    
2474      /* In the real compile phase, just check the workspace used by the forward
2475      reference list. */
2476    
2477      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2478        {
2479        *errorcodeptr = ERR52;
2480        goto FAILED;
2481        }
2482    
2483    /* 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 */
2484    
2485    if (inescq && c != 0)    if (inescq && c != 0)
# Line 1623  for (;; ptr++) Line 2494  for (;; ptr++)
2494        {        {
2495        if (previous_callout != NULL)        if (previous_callout != NULL)
2496          {          {
2497          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2498              complete_callout(previous_callout, ptr, cd);
2499          previous_callout = NULL;          previous_callout = NULL;
2500          }          }
2501        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1644  for (;; ptr++) Line 2516  for (;; ptr++)
2516    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2517         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2518      {      {
2519      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2520          complete_callout(previous_callout, ptr, cd);
2521      previous_callout = NULL;      previous_callout = NULL;
2522      }      }
2523    
# Line 1655  for (;; ptr++) Line 2528  for (;; ptr++)
2528      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2529      if (c == '#')      if (c == '#')
2530        {        {
2531        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2532        on the Macintosh. */          {
2533        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2534        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2535          if (*ptr != 0) continue;
2536    
2537          /* Else fall through to handle end of string */
2538          c = 0;
2539        }        }
2540      }      }
2541    
# Line 1672  for (;; ptr++) Line 2549  for (;; ptr++)
2549    
2550    switch(c)    switch(c)
2551      {      {
2552      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2553        case 0:                        /* The branch terminates at string end */
2554      case 0:      case '|':                      /* or | or ) */
     case '|':  
2555      case ')':      case ')':
2556      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2557      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2558      *codeptr = code;      *codeptr = code;
2559      *ptrptr = ptr;      *ptrptr = ptr;
2560        if (lengthptr != NULL)
2561          {
2562          if (OFLOW_MAX - *lengthptr < code - last_code)
2563            {
2564            *errorcodeptr = ERR20;
2565            goto FAILED;
2566            }
2567          *lengthptr += code - last_code;   /* To include callout length */
2568          DPRINTF((">> end branch\n"));
2569          }
2570      return TRUE;      return TRUE;
2571    
2572    
2573        /* ===================================================================*/
2574      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2575      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2576    
# Line 1711  for (;; ptr++) Line 2599  for (;; ptr++)
2599      *code++ = OP_ANY;      *code++ = OP_ANY;
2600      break;      break;
2601    
2602      /* Character classes. If the included characters are all < 255 in value, we  
2603      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2604      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
2605      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
2606      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2607        map as usual, then invert it at the end. However, we use a different opcode
2608        so that data characters > 255 can be handled correctly.
2609    
2610      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2611      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
# Line 1736  for (;; ptr++) Line 2626  for (;; ptr++)
2626        goto FAILED;        goto FAILED;
2627        }        }
2628    
2629      /* 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,
2630        if the first few characters (either before or after ^) are \Q\E or \E we
2631        skip them too. This makes for compatibility with Perl. */
2632    
2633      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2634        for (;;)
2635        {        {
       negate_class = TRUE;  
2636        c = *(++ptr);        c = *(++ptr);
2637        }        if (c == '\\')
2638      else          {
2639        {          if (ptr[1] == 'E') ptr++;
2640        negate_class = FALSE;            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2641                else break;
2642            }
2643          else if (!negate_class && c == '^')
2644            negate_class = TRUE;
2645          else break;
2646        }        }
2647    
2648        /* If a class contains a negative special such as \S, we need to flip the
2649        negation flag at the end, so that support for characters > 255 works
2650        correctly (they are all included in the class). */
2651    
2652        should_flip_negation = FALSE;
2653    
2654      /* 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
2655      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
2656      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2657    
2658      class_charcount = 0;      class_charcount = 0;
2659      class_lastchar = -1;      class_lastchar = -1;
2660    
2661        /* Initialize the 32-char bit map to all zeros. We build the map in a
2662        temporary bit of memory, in case the class contains only 1 character (less
2663        than 256), because in that case the compiled code doesn't use the bit map.
2664        */
2665    
2666        memset(classbits, 0, 32 * sizeof(uschar));
2667    
2668  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2669      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2670      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2671  #endif  #endif
2672    
     /* 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));  
   
2673      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2674      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
2675      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. */  
2676    
2677      do      if (c != 0) do
2678        {        {
2679          const uschar *oldptr;
2680    
2681  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2682        if (utf8 && c > 127)        if (utf8 && c > 127)
2683          {                           /* Braces are required because the */          {                           /* Braces are required because the */
# Line 1786  for (;; ptr++) Line 2689  for (;; ptr++)
2689    
2690        if (inescq)        if (inescq)
2691          {          {
2692          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2693            {            {
2694            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2695            ptr++;            ptr++;                            /* Skip the 'E' */
2696            continue;            continue;                         /* Carry on with next */
2697            }            }
2698          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2699          }          }
2700    
2701        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1806  for (;; ptr++) Line 2709  for (;; ptr++)
2709            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr, cd))
2710          {          {
2711          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2712          int posix_class, i;          int posix_class, taboffset, tabopt;
2713          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2714            uschar pbits[32];
2715    
2716          if (ptr[1] != ':')          if (ptr[1] != ':')
2717            {            {
# Line 1819  for (;; ptr++) Line 2723  for (;; ptr++)
2723          if (*ptr == '^')          if (*ptr == '^')
2724            {            {
2725            local_negate = TRUE;            local_negate = TRUE;
2726              should_flip_negation = TRUE;  /* Note negative special */
2727            ptr++;            ptr++;
2728            }            }
2729    
# Line 1836  for (;; ptr++) Line 2741  for (;; ptr++)
2741          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
2742            posix_class = 0;            posix_class = 0;
2743    
2744          /* 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
2745          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
2746          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
2747          white space chars afterwards. */          result into the bit map that is being built. */
2748    
2749          posix_class *= 3;          posix_class *= 3;
2750          for (i = 0; i < 3; i++)  
2751            /* Copy in the first table (always present) */
2752    
2753            memcpy(pbits, cbits + posix_class_maps[posix_class],
2754              32 * sizeof(uschar));
2755    
2756            /* If there is a second table, add or remove it as required. */
2757    
2758            taboffset = posix_class_maps[posix_class + 1];
2759            tabopt = posix_class_maps[posix_class + 2];
2760    
2761            if (taboffset >= 0)
2762            {            {
2763            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
2764            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;  
             }  
2765            else            else
2766              {              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;  
             }  
2767            }            }
2768    
2769            /* Not see if we need to remove any special characters. An option
2770            value of 1 removes vertical space and 2 removes underscore. */
2771    
2772            if (tabopt < 0) tabopt = -tabopt;
2773            if (tabopt == 1) pbits[1] &= ~0x3c;
2774              else if (tabopt == 2) pbits[11] &= 0x7f;
2775    
2776            /* Add the POSIX table or its complement into the main table that is
2777            being built and we are done. */
2778    
2779            if (local_negate)
2780              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
2781            else
2782              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
2783    
2784          ptr = tempptr + 1;          ptr = tempptr + 1;
2785          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
2786          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
2787          }          }
2788    
2789        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2790        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
2791        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.
2792        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2793        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  
2794        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2795    
2796        if (c == '\\')        if (c == '\\')
2797          {          {
2798          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2799            if (*errorcodeptr != 0) goto FAILED;
2800    
2801          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2802          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 */
2803            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2804          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2805            {            {
2806            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1890  for (;; ptr++) Line 2810  for (;; ptr++)
2810            else inescq = TRUE;            else inescq = TRUE;
2811            continue;            continue;
2812            }            }
2813            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2814    
2815          if (c < 0)          if (c < 0)
2816            {            {
2817            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2818            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2819            switch (-c)  
2820              /* Save time by not doing this in the pre-compile phase. */
2821    
2822              if (lengthptr == NULL) switch (-c)
2823              {              {
2824              case ESC_d:              case ESC_d:
2825              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
2826              continue;              continue;
2827    
2828              case ESC_D:              case ESC_D:
2829                should_flip_negation = TRUE;
2830              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2831              continue;              continue;
2832    
# Line 1910  for (;; ptr++) Line 2835  for (;; ptr++)
2835              continue;              continue;
2836    
2837              case ESC_W:              case ESC_W:
2838                should_flip_negation = TRUE;
2839              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2840              continue;              continue;
2841    
# Line 1919  for (;; ptr++) Line 2845  for (;; ptr++)
2845              continue;              continue;
2846    
2847              case ESC_S:              case ESC_S:
2848                should_flip_negation = TRUE;
2849              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2850              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2851              continue;              continue;
2852    
2853  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
2854              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
2855              case ESC_P:              }
2856    
2857              /* In the pre-compile phase, just do the recognition. */
2858    
2859              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2860                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
2861    
2862              /* We need to deal with \H, \h, \V, and \v in both phases because
2863              they use extra memory. */
2864    
2865              if (-c == ESC_h)
2866                {
2867                SETBIT(classbits, 0x09); /* VT */
2868                SETBIT(classbits, 0x20); /* SPACE */
2869                SETBIT(classbits, 0xa0); /* NSBP */
2870    #ifdef SUPPORT_UTF8
2871                if (utf8)
2872                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
2873                class_utf8 = TRUE;                class_utf8 = TRUE;
2874                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_SINGLE;
2875                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2876                *class_utf8data++ = property;                *class_utf8data++ = XCL_SINGLE;
2877                class_charcount -= 2;   /* Not a < 256 character */                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2878                  *class_utf8data++ = XCL_RANGE;
2879                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2880                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2881                  *class_utf8data++ = XCL_SINGLE;
2882                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2883                  *class_utf8data++ = XCL_SINGLE;
2884                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2885                  *class_utf8data++ = XCL_SINGLE;
2886                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2887                }                }
             continue;  
2888  #endif  #endif
2889                continue;
2890                }
2891    
2892              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_H)
2893              strict mode. By default, for compatibility with Perl, they are              {
2894              treated as literals. */              for (c = 0; c < 32; c++)
2895                  {
2896                  int x = 0xff;
2897                  switch (c)
2898                    {
2899                    case 0x09/8: x ^= 1 << (0x09%8); break;
2900                    case 0x20/8: x ^= 1 << (0x20%8); break;
2901                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2902                    default: break;
2903                    }
2904                  classbits[c] |= x;
2905                  }
2906    
2907              default:  #ifdef SUPPORT_UTF8
2908              if ((options & PCRE_EXTRA) != 0)              if (utf8)
2909                {                {
2910                *errorcodeptr = ERR7;                class_utf8 = TRUE;
2911                goto FAILED;                *class_utf8data++ = XCL_RANGE;
2912                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2913                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
2914                  *class_utf8data++ = XCL_RANGE;
2915                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
2916                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
2917                  *class_utf8data++ = XCL_RANGE;
2918                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
2919                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
2920                  *class_utf8data++ = XCL_RANGE;
2921                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
2922                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
2923                  *class_utf8data++ = XCL_RANGE;
2924                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
2925                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
2926                  *class_utf8data++ = XCL_RANGE;
2927                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
2928                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
2929                  *class_utf8data++ = XCL_RANGE;
2930                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
2931                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2932                }                }
2933              c = *ptr;              /* The final character */  #endif
2934              class_charcount -= 2;  /* Undo the default count from above */              continue;
2935              }              }
           }  
2936    
2937          /* Fall through if we have a single character (c >= 0). This may be            if (-c == ESC_v)
2938          > 256 in UTF-8 mode. */              {
2939                SETBIT(classbits, 0x0a); /* LF */
2940                SETBIT(classbits, 0x0b); /* VT */
2941                SETBIT(classbits, 0x0c); /* FF */
2942                SETBIT(classbits, 0x0d); /* CR */
2943                SETBIT(classbits, 0x85); /* NEL */
2944    #ifdef SUPPORT_UTF8
2945                if (utf8)
2946                  {
2947                  class_utf8 = TRUE;
2948                  *class_utf8data++ = XCL_RANGE;
2949                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
2950                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2951                  }
2952    #endif
2953                continue;
2954                }
2955    
2956          }   /* End of backslash handling */            if (-c == ESC_V)
2957                {
2958                for (c = 0; c < 32; c++)
2959                  {
2960                  int x = 0xff;
2961                  switch (c)
2962                    {
2963                    case 0x0a/8: x ^= 1 << (0x0a%8);
2964                                 x ^= 1 << (0x0b%8);
2965                                 x ^= 1 << (0x0c%8);
2966                                 x ^= 1 << (0x0d%8);
2967                                 break;
2968                    case 0x85/8: x ^= 1 << (0x85%8); break;
2969                    default: break;
2970                    }
2971                  classbits[c] |= x;
2972                  }
2973    
2974    #ifdef SUPPORT_UTF8
2975                if (utf8)
2976                  {
2977                  class_utf8 = TRUE;
2978                  *class_utf8data++ = XCL_RANGE;
2979                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2980                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
2981                  *class_utf8data++ = XCL_RANGE;
2982                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2983                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2984                  }
2985    #endif
2986                continue;
2987                }
2988    
2989              /* We need to deal with \P and \p in both phases. */
2990    
2991    #ifdef SUPPORT_UCP
2992              if (-c == ESC_p || -c == ESC_P)
2993                {
2994                BOOL negated;
2995                int pdata;
2996                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
2997                if (ptype < 0) goto FAILED;
2998                class_utf8 = TRUE;
2999                *class_utf8data++ = ((-c == ESC_p) != negated)?
3000                  XCL_PROP : XCL_NOTPROP;
3001                *class_utf8data++ = ptype;
3002                *class_utf8data++ = pdata;
3003                class_charcount -= 2;   /* Not a < 256 character */
3004                continue;
3005                }
3006    #endif
3007              /* Unrecognized escapes are faulted if PCRE is running in its
3008              strict mode. By default, for compatibility with Perl, they are
3009              treated as literals. */
3010    
3011              if ((options & PCRE_EXTRA) != 0)
3012                {
3013                *errorcodeptr = ERR7;
3014                goto FAILED;
3015                }
3016    
3017              class_charcount -= 2;  /* Undo the default count from above */
3018              c = *ptr;              /* Get the final character and fall through */
3019              }
3020    
3021            /* Fall through if we have a single character (c >= 0). This may be
3022            greater than 256 in UTF-8 mode. */
3023    
3024            }   /* End of backslash handling */
3025    
3026        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3027        Perl does not permit ']' to be the end of the range. A '-' character        Perl does not permit ']' to be the end of the range. A '-' character
3028        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3029          entirely. The code for handling \Q and \E is messy. */
3030    
3031          CHECK_RANGE:
3032          while (ptr[1] == '\\' && ptr[2] == 'E')
3033            {
3034            inescq = FALSE;
3035            ptr += 2;
3036            }
3037    
3038        if (ptr[1] == '-' && ptr[2] != ']')        oldptr = ptr;
3039    
3040          /* Remember \r or \n */
3041    
3042          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3043    
3044          /* Check for range */
3045    
3046          if (!inescq && ptr[1] == '-')
3047          {          {
3048          int d;          int d;
3049          ptr += 2;          ptr += 2;
3050            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
3051    
3052            /* If we hit \Q (not followed by \E) at this point, go into escaped
3053            mode. */
3054    
3055            while (*ptr == '\\' && ptr[1] == 'Q')
3056              {
3057              ptr += 2;
3058              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
3059              inescq = TRUE;
3060              break;
3061              }
3062    
3063            if (*ptr == 0 || (!inescq && *ptr == ']'))
3064              {
3065              ptr = oldptr;
3066              goto LONE_SINGLE_CHARACTER;
3067              }
3068    
3069  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3070          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3079  for (;; ptr++)
3079          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
3080          in such circumstances. */          in such circumstances. */
3081    
3082          if (d == '\\')          if (!inescq && d == '\\')
3083            {            {
3084            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3085            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3086    
3087            /* \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
3088            was literal */            special means the '-' was literal */
3089    
3090            if (d < 0)            if (d < 0)
3091              {              {
3092              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = '\b';
3093              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = 'X';
3094                else if (d == -ESC_R) d = 'R'; else
3095                {                {
3096                ptr = oldptr - 2;                ptr = oldptr;
3097                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3098                }                }
3099              }              }
3100            }            }
3101    
3102          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3103          the pre-pass. Optimize one-character ranges */          one-character ranges */
3104    
3105            if (d < c)
3106              {
3107              *errorcodeptr = ERR8;
3108              goto FAILED;
3109              }
3110    
3111          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3112    
3113            /* Remember \r or \n */
3114    
3115            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3116    
3117          /* 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
3118          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3119          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 3131  for (;; ptr++)
3131  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3132            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3133              {              {
3134              int occ, ocd;              unsigned int occ, ocd;
3135              int cc = c;              unsigned int cc = c;
3136              int origd = d;              unsigned int origd = d;
3137              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3138                {                {
3139                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3140                      ocd <= (unsigned int)d)
3141                    continue;                          /* Skip embedded ranges */
3142    
3143                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3144                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3145                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3146                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3147                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3148                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3149                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3150                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3151                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3152                  d = ocd;                  d = ocd;
3153                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3195  for (;; ptr++)
3195          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
3196          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3197    
3198          for (; c <= d; c++)          class_charcount += d - c + 1;
3199            class_lastchar = d;
3200    
3201            /* We can save a bit of time by skipping this in the pre-compile. */
3202    
3203            if (lengthptr == NULL) for (; c <= d; c++)
3204            {            {
3205            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3206            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3208  for (;; ptr++)
3208              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3209              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3210              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3211            }            }
3212    
3213          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3231  for (;; ptr++)
3231  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3232          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3233            {            {
3234            int chartype;            unsigned int othercase;
3235            int othercase;            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3236              {              {
3237              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3238              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3257  for (;; ptr++)
3257          }          }
3258        }        }
3259    
3260      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3261      loop. This "while" is the end of the "do" above. */  
3262        while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
3263    
3264        if (c == 0)                          /* Missing terminating ']' */
3265          {
3266          *errorcodeptr = ERR6;
3267          goto FAILED;
3268          }
3269    
3270    
3271    /* This code has been disabled because it would mean that \s counts as
3272    an explicit \r or \n reference, and that's not really what is wanted. Now
3273    we set the flag only if there is a literal "\r" or "\n" in the class. */
3274    
3275    #if 0
3276        /* Remember whether \r or \n are in this class */
3277    
3278        if (negate_class)
3279          {
3280          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3281          }
3282        else
3283          {
3284          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3285          }
3286    #endif
3287    
     while ((c = *(++ptr)) != ']' || inescq);  
3288    
3289      /* 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
3290      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
3291      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
3292      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3293      single-bytes only. This is an historical hangover. Maybe one day we can  
3294      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3295        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3296        operate on single-bytes only. This is an historical hangover. Maybe one day
3297        we can tidy these opcodes to handle multi-byte characters.
3298    
3299      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
3300      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 3304  for (;; ptr++)
3304      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3305    
3306  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3307      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3308            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3309  #else  #else
3310      if (class_charcount == 1)      if (class_charcount == 1)
3311  #endif  #endif
# Line 2209  for (;; ptr++) Line 3348  for (;; ptr++)
3348      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3349    
3350      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3351      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3352      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3353        the class, so any that were explicitly given as well can be ignored. If
3354        (when there are explicit characters > 255 that must be listed) there are no
3355        characters < 256, we can omit the bitmap in the actual compiled code. */
3356    
3357  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3358      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3359        {        {
3360        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3361        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3362        code += LINK_SIZE;        code += LINK_SIZE;
3363        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3364    
3365        /* 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;
3366        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3367    
3368        if (class_charcount > 0)        if (class_charcount > 0)
3369          {          {
3370          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3371            memmove(code + 32, code, class_utf8data - code);
3372          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3373          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;  
3374          }          }
3375          else code = class_utf8data;
3376    
3377        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3378    
# Line 2246  for (;; ptr++) Line 3381  for (;; ptr++)
3381        }        }
3382  #endif  #endif
3383    
3384      /* 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
3385      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
3386      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
3387      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3388    
3389        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3390      if (negate_class)      if (negate_class)
3391        {        {
3392        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3393        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3394        }        }
3395      else      else
3396        {        {
       *code++ = OP_CLASS;  
3397        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3398        }        }
3399      code += 32;      code += 32;
3400      break;      break;
3401    
3402    
3403        /* ===================================================================*/
3404      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3405      has been tested above. */      has been tested above. */
3406    
# Line 2331  for (;; ptr++) Line 3468  for (;; ptr++)
3468        }        }
3469      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3470    
     /* 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;  
       }  
   
3471      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3472      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
3473      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 3501  for (;; ptr++)
3501          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3502          }          }
3503    
3504          /* If the repetition is unlimited, it pays to see if the next thing on
3505          the line is something that cannot possibly match this character. If so,
3506          automatically possessifying this item gains some performance in the case
3507          where the match fails. */
3508    
3509          if (!possessive_quantifier &&
3510              repeat_max < 0 &&
3511              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3512                options, cd))
3513            {
3514            repeat_type = 0;    /* Force greedy */
3515            possessive_quantifier = TRUE;
3516            }
3517    
3518        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3519        }        }
3520    
3521      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3522      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-
3523      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3524      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3525        currently used only for single-byte chars. */
3526    
3527      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3528        {        {
3529        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3530        c = previous[1];        c = previous[1];
3531          if (!possessive_quantifier &&
3532              repeat_max < 0 &&
3533              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3534            {
3535            repeat_type = 0;    /* Force greedy */
3536            possessive_quantifier = TRUE;
3537            }
3538        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3539        }        }
3540    
# Line 2403  for (;; ptr++) Line 3548  for (;; ptr++)
3548      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3549        {        {
3550        uschar *oldcode;        uschar *oldcode;
3551        int prop_type;        int prop_type, prop_value;
3552        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3553        c = *previous;        c = *previous;
3554    
3555          if (!possessive_quantifier &&
3556              repeat_max < 0 &&
3557              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3558            {
3559            repeat_type = 0;    /* Force greedy */
3560            possessive_quantifier = TRUE;
3561            }
3562    
3563        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3564        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3565          previous[1] : -1;          {
3566            prop_type = previous[1];
3567            prop_value = previous[2];
3568            }
3569          else prop_type = prop_value = -1;
3570    
3571        oldcode = code;        oldcode = code;
3572        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2422  for (;; ptr++) Line 3579  for (;; ptr++)
3579        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3580        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3581    
3582        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3583    
3584        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3585    
# Line 2443  for (;; ptr++) Line 3600  for (;; ptr++)
3600          }          }
3601    
3602        /* 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
3603        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3604        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
3605        one less than the maximum. */        one less than the maximum. */
3606    
# Line 2470  for (;; ptr++) Line 3627  for (;; ptr++)
3627    
3628          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3629          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
3630          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3631          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3632          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3633    
# Line 2486  for (;; ptr++) Line 3643  for (;; ptr++)
3643  #endif  #endif
3644              {              {
3645              *code++ = c;              *code++ = c;
3646              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3647                  {
3648                  *code++ = prop_type;
3649                  *code++ = prop_value;
3650                  }
3651              }              }
3652            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3653            }            }
3654    
3655          /* 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
3656          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3657            UPTO is just for 1 instance, we can use QUERY instead. */
3658    
3659          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3660            {            {
# Line 2505  for (;; ptr++) Line 3667  for (;; ptr++)
3667            else            else
3668  #endif  #endif
3669            *code++ = c;            *code++ = c;
3670            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3671                {
3672                *code++ = prop_type;
3673                *code++ = prop_value;
3674                }
3675            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3676            *code++ = OP_UPTO + repeat_type;  
3677            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3678                {
3679                *code++ = OP_QUERY + repeat_type;
3680                }
3681              else
3682                {
3683                *code++ = OP_UPTO + repeat_type;
3684                PUT2INC(code, 0, repeat_max);
3685                }
3686            }            }
3687          }          }
3688    
# Line 2524  for (;; ptr++) Line 3698  for (;; ptr++)
3698  #endif  #endif
3699        *code++ = c;        *code++ = c;
3700    
3701        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3702        defines the required property. */        define the required property. */
3703    
3704  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3705        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3706            {
3707            *code++ = prop_type;
3708            *code++ = prop_value;
3709            }
3710  #endif  #endif
3711        }        }
3712    
# Line 2551  for (;; ptr++) Line 3729  for (;; ptr++)
3729        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3730        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3731    
3732        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3733    
3734        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3735          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2571  for (;; ptr++) Line 3749  for (;; ptr++)
3749      /* 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
3750      cases. */      cases. */
3751    
3752      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3753               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3754        {        {
3755        register int i;        register int i;
3756        int ketoffset = 0;        int ketoffset = 0;
3757        int len = code - previous;        int len = code - previous;
3758        uschar *bralink = NULL;        uschar *bralink = NULL;
3759    
3760          /* Repeating a DEFINE group is pointless */
3761    
3762          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3763            {
3764            *errorcodeptr = ERR55;
3765            goto FAILED;
3766            }
3767    
3768        /* 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
3769        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
3770        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 3799  for (;; ptr++)
3799          /* 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
3800          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
3801          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
3802          any internal group, because the offset is from the start of the whole          any internal or forward referenced group, because the offset is from
3803          regex. Temporarily terminate the pattern while doing this. */          the start of the whole regex. Temporarily terminate the pattern while
3804            doing this. */
3805    
3806          if (repeat_max <= 1)          if (repeat_max <= 1)
3807            {            {
3808            *code = OP_END;            *code = OP_END;
3809            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3810            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3811            code++;            code++;
3812            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2637  for (;; ptr++) Line 3824  for (;; ptr++)
3824            {            {
3825            int offset;            int offset;
3826            *code = OP_END;            *code = OP_END;
3827            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3828            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3829            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3830            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 3844  for (;; ptr++)
3844        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3845        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3846        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
3847        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3848          forward reference subroutine calls in the group, there will be entries on
3849          the workspace list; replicate these with an appropriate increment. */
3850    
3851        else        else
3852          {          {
3853          if (repeat_min > 1)          if (repeat_min > 1)
3854            {            {
3855            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3856            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
3857              potential integer overflow. */
3858    
3859              if (lengthptr != NULL)
3860              {              {
3861              memcpy(code, previous, len);              int delta = (repeat_min - 1)*length_prevgroup;
3862              code += len;              if ((double)(repeat_min - 1)*(double)length_prevgroup >
3863                                                                (double)INT_MAX ||
3864                    OFLOW_MAX - *lengthptr < delta)
3865                  {
3866                  *errorcodeptr = ERR20;
3867                  goto FAILED;
3868                  }
3869                *lengthptr += delta;
3870                }
3871    
3872              /* This is compiling for real */
3873    
3874              else
3875                {
3876                if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3877                for (i = 1; i < repeat_min; i++)
3878                  {
3879                  uschar *hc;
3880                  uschar *this_hwm = cd->hwm;
3881                  memcpy(code, previous, len);
3882                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3883                    {
3884                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3885                    cd->hwm += LINK_SIZE;
3886                    }
3887                  save_hwm = this_hwm;
3888                  code += len;
3889                  }
3890              }              }
3891            }            }
3892    
3893          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3894          }          }
3895    
# Line 2677  for (;; ptr++) Line 3897  for (;; ptr++)
3897        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3898        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,
3899        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
3900        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
3901          replicate entries on the forward reference list. */
3902    
3903        if (repeat_max >= 0)        if (repeat_max >= 0)
3904          {          {
3905          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
3906            just adjust the length as if we had. For each repetition we must add 1
3907            to the length for BRAZERO and for all but the last repetition we must
3908            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3909            paranoid checks to avoid integer overflow. */
3910    
3911            if (lengthptr != NULL && repeat_max > 0)
3912              {
3913              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3914                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3915              if ((double)repeat_max *
3916                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3917                      > (double)INT_MAX ||
3918                  OFLOW_MAX - *lengthptr < delta)
3919                {
3920                *errorcodeptr = ERR20;
3921                goto FAILED;
3922                }
3923              *lengthptr += delta;
3924              }
3925    
3926            /* This is compiling for real */
3927    
3928            else for (i = repeat_max - 1; i >= 0; i--)
3929            {            {
3930              uschar *hc;
3931              uschar *this_hwm = cd->hwm;
3932    
3933            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
3934    
3935            /* 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 3945  for (;; ptr++)
3945              }              }
3946    
3947            memcpy(code, previous, len);            memcpy(code, previous, len);
3948              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3949                {
3950                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
3951                cd->hwm += LINK_SIZE;
3952                }
3953              save_hwm = this_hwm;
3954            code += len;            code += len;
3955            }            }
3956    
# Line 2720  for (;; ptr++) Line 3973  for (;; ptr++)
3973        /* 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
3974        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
3975        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
3976        correct offset was computed above. */        correct offset was computed above.
3977    
3978        else code[-ketoffset] = OP_KETRMAX + repeat_type;        Then, when we are doing the actual compile phase, check to see whether
3979          this group is a non-atomic one that could match an empty string. If so,
3980          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
3981          that runtime checking can be done. [This check is also applied to
3982          atomic groups at runtime, but in a different way.] */
3983    
3984          else
3985            {
3986            uschar *ketcode = code - ketoffset;
3987            uschar *bracode = ketcode - GET(ketcode, 1);
3988            *ketcode = OP_KETRMAX + repeat_type;
3989            if (lengthptr == NULL && *bracode != OP_ONCE)
3990              {
3991              uschar *scode = bracode;
3992              do
3993                {
3994                if (could_be_empty_branch(scode, ketcode, utf8))
3995                  {
3996                  *bracode += OP_SBRA - OP_BRA;
3997                  break;
3998                  }
3999                scode += GET(scode, 1);
4000                }
4001              while (*scode == OP_ALT);
4002              }
4003            }
4004        }        }
4005    
4006      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2733  for (;; ptr++) Line 4011  for (;; ptr++)
4011        goto FAILED;        goto FAILED;
4012        }        }
4013    
4014      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
4015      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
4016      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
4017      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.
4018      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
4019        but the special opcodes can optimize it a bit. The repeated item starts at
4020        tempcode, not at previous, which might be the first part of a string whose
4021        (former) last char we repeated.
4022    
4023        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4024        an 'upto' may follow. We skip over an 'exact' item, and then test the
4025        length of what remains before proceeding. */
4026    
4027      if (possessive_quantifier)      if (possessive_quantifier)
4028        {        {
4029        int len = code - tempcode;        int len;
4030        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4031        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
4032        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode] +
4033        tempcode[0] = OP_ONCE;            ((*tempcode == OP_TYPEEXACT &&
4034        *code++ = OP_KET;               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4035        PUTINC(code, 0, len);        len = code - tempcode;
4036        PUT(tempcode, 1, len);        if (len > 0) switch (*tempcode)
4037            {
4038            case OP_STAR:  *tempcode = OP_POSSTAR; break;
4039            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
4040            case OP_QUERY: *tempcode = OP_POSQUERY; break;
4041            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4042    
4043            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4044            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4045            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4046            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4047    
4048            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4049            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4050            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4051            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4052    
4053            default:
4054            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4055            code += 1 + LINK_SIZE;
4056            len += 1 + LINK_SIZE;
4057            tempcode[0] = OP_ONCE;
4058            *code++ = OP_KET;
4059            PUTINC(code, 0, len);
4060            PUT(tempcode, 1, len);
4061            break;
4062            }
4063        }        }
4064    
4065      /* 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 4072  for (;; ptr++)
4072      break;      break;
4073    
4074    
4075      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
4076      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4077      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
4078      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.  */  
4079    
4080      case '(':      case '(':
4081      newoptions = options;      newoptions = options;
4082      skipbytes = 0;      skipbytes = 0;
4083        bravalue = OP_CBRA;
4084        save_hwm = cd->hwm;
4085        reset_bracount = FALSE;
4086    
4087        /* First deal with various "verbs" that can be introduced by '*'. */
4088    
4089        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4090          {
4091          int i, namelen;
4092          const char *vn = verbnames;
4093          const uschar *name = ++ptr;
4094          previous = NULL;
4095          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4096          if (*ptr == ':')
4097            {
4098            *errorcodeptr = ERR59;   /* Not supported */
4099            goto FAILED;
4100            }
4101          if (*ptr != ')')
4102            {
4103            *errorcodeptr = ERR60;
4104            goto FAILED;
4105            }
4106          namelen = ptr - name;
4107          for (i = 0; i < verbcount; i++)
4108            {
4109            if (namelen == verbs[i].len &&
4110                strncmp((char *)name, vn, namelen) == 0)
4111              {
4112              *code = verbs[i].op;
4113              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4114              break;
4115              }
4116            vn += verbs[i].len + 1;
4117            }
4118          if (i < verbcount) continue;
4119          *errorcodeptr = ERR60;
4120          goto FAILED;
4121          }
4122    
4123        /* Deal with the extended parentheses; all are introduced by '?', and the
4124        appearance of any of them means that this is not a capturing group. */
4125    
4126      if (*(++ptr) == '?')      else if (*ptr == '?')
4127        {        {
4128        int set, unset;        int i, set, unset, namelen;
4129        int *optset;        int *optset;
4130          const uschar *name;
4131          uschar *slot;
4132    
4133        switch (*(++ptr))        switch (*(++ptr))
4134          {          {
4135          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
4136          ptr++;          ptr++;
4137          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
4138            if (*ptr == 0)
4139              {
4140              *errorcodeptr = ERR18;
4141              goto FAILED;
4142              }
4143          continue;          continue;
4144    
4145          case ':':                 /* Non-extracting bracket */  
4146            /* ------------------------------------------------------------ */
4147            case '|':                 /* Reset capture count for each branch */
4148            reset_bracount = TRUE;
4149            /* Fall through */
4150    
4151            /* ------------------------------------------------------------ */
4152            case ':':                 /* Non-capturing bracket */
4153          bravalue = OP_BRA;          bravalue = OP_BRA;
4154          ptr++;          ptr++;
4155          break;          break;
4156    
4157    
4158            /* ------------------------------------------------------------ */
4159          case '(':          case '(':
4160          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4161    
4162          /* Condition to test for recursion */          /* A condition can be an assertion, a number (referring to a numbered
4163            group), a name (referring to a named group), or 'R', referring to
4164            recursion. R<digits> and R&name are also permitted for recursion tests.
4165    
4166            There are several syntaxes for testing a named group: (?(name)) is used
4167            by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
4168    
4169            There are two unfortunate ambiguities, caused by history. (a) 'R' can
4170            be the recursive thing or the name 'R' (and similarly for 'R' followed
4171            by digits), and (b) a number could be a name that consists of digits.
4172            In both cases, we look for a name first; if not found, we try the other
4173            cases. */
4174    
4175            /* For conditions that are assertions, check the syntax, and then exit
4176            the switch. This will take control down to where bracketed groups,
4177            including assertions, are processed. */
4178    
4179            if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
4180              break;
4181    
4182            /* Most other conditions use OP_CREF (a couple change to OP_RREF
4183            below), and all need to skip 3 bytes at the start of the group. */
4184    
4185          if (ptr[1] == 'R')          code[1+LINK_SIZE] = OP_CREF;
4186            skipbytes = 3;
4187            refsign = -1;
4188    
4189            /* Check for a test for recursion in a named group. */
4190    
4191            if (ptr[1] == 'R' && ptr[2] == '&')
4192            {            {
4193            code[1+LINK_SIZE] = OP_CREF;            terminator = -1;
4194            PUT2(code, 2+LINK_SIZE, CREF_RECURSE);            ptr += 2;
4195            skipbytes = 3;            code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
           ptr += 3;  
4196            }            }
4197    
4198          /* 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
4199          if a digit follows ( then there will just be digits until ) because          syntax (?(<name>) or (?('name') */
         the syntax was checked in the first pass. */  
4200    
4201          else if ((digitab[ptr[1]] && ctype_digit) != 0)          else if (ptr[1] == '<')
4202            {            {
4203            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;  
             }  
4204            ptr++;            ptr++;
           code[1+LINK_SIZE] = OP_CREF;  
           PUT2(code, 2+LINK_SIZE, condref);  
           skipbytes = 3;  
4205            }            }
4206          /* 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))  
4207            {            {
4208            case '=':               /* Positive lookbehind */            terminator = '\'';
           bravalue = OP_ASSERTBACK;  
4209            ptr++;            ptr++;
4210            break;            }
4211            else
4212              {
4213              terminator = 0;
4214              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4215              }
4216    
4217            case '!':               /* Negative lookbehind */          /* We now expect to read a name; any thing else is an error */
4218            bravalue = OP_ASSERTBACK_NOT;  
4219            ptr++;          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
4220            break;            {
4221              ptr += 1;  /* To get the right offset */
4222              *errorcodeptr = ERR28;
4223              goto FAILED;
4224            }            }
         break;  
4225    
4226          case '>':                 /* One-time brackets */          /* Read the name, but also get it as a number if it's all digits */
         bravalue = OP_ONCE;  
         ptr++;  
         break;  
4227    
4228          case 'C':                 /* Callout - may be followed by digits; */          recno = 0;
4229          previous_callout = code;  /* Save for later completion */          name = ++ptr;
4230          after_manual_callout = 1; /* Skip one item before completing */          while ((cd->ctypes[*ptr] & ctype_word) != 0)
4231          *code++ = OP_CALLOUT;     /* Already checked that the terminating */            {
4232            {                       /* closing parenthesis is present. */            if (recno >= 0)
4233            int n = 0;              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4234            while ((digitab[*(++ptr)] & ctype_digit) != 0)                recno * 10 + *ptr - '0' : -1;
4235              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;  
4236            }            }
4237          previous = NULL;          namelen = ptr - name;
         continue;  
4238    
4239          case 'P':                 /* Named subpattern handling */          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
         if (*(++ptr) == '<')      /* Definition */  
4240            {            {
4241            int i, namelen;            ptr--;      /* Error offset */
4242            uschar *slot = cd->name_table;            *errorcodeptr = ERR26;
4243            const uschar *name;     /* Don't amalgamate; some compilers */            goto FAILED;
4244            name = ++ptr;           /* grumble at autoincrement in declaration */            }
4245    
4246            while (*ptr++ != '>');          /* Do no further checking in the pre-compile phase. */
           namelen = ptr - name - 1;  
4247    
4248            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;  
             }  
4249    
4250            PUT2(slot, 0, *brackets + 1);          /* In the real compile we do the work of looking for the actual
4251            memcpy(slot + 2, name, namelen);          reference. If the string started with "+" or "-" we require the rest to
4252            slot[2+namelen] = 0;          be digits, in which case recno will be set. */
           cd->names_found++;  
           goto NUMBERED_GROUP;  
           }  
4253    
4254          if (*ptr == '=' || *ptr == '>')  /* Reference or recursion */          if (refsign > 0)
4255            {            {
4256            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++)  
4257              {              {
4258              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              *errorcodeptr = ERR58;
4259              slot += cd->name_entry_size;              goto FAILED;
4260              }              }
4261            if (i >= cd->names_found)            recno = (refsign == '-')?
4262                cd->bracount - recno + 1 : recno +cd->bracount;
4263              if (recno <= 0 || recno > cd->final_bracount)
4264              {              {
4265              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
4266              goto FAILED;              goto FAILED;
4267              }              }
4268              PUT2(code, 2+LINK_SIZE, recno);
4269              break;
4270              }
4271    
4272            recno = GET2(slot, 0);          /* Otherwise (did not start with "+" or "-"), start by looking for the
4273            name. */
           if (type == '>') goto HANDLE_RECURSION;  /* A few lines below */  
4274    
4275            /* Back reference */          slot = cd->name_table;
4276            for (i = 0; i < cd->names_found; i++)
4277              {
4278              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
4279              slot += cd->name_entry_size;
4280              }
4281    
4282            previous = code;          /* Found a previous named subpattern */
4283            *code++ = OP_REF;  
4284            PUT2INC(code, 0, recno);          if (i < cd->names_found)
4285            cd->backref_map |= (recno < 32)? (1 << recno) : 1;            {
4286            if (recno > cd->top_backref) cd->top_backref = recno;            recno = GET2(slot, 0);
4287              PUT2(code, 2+LINK_SIZE, recno);
4288              }
4289    
4290            /* Search the pattern for a forward reference */
4291    
4292            else if ((i = find_parens(ptr, cd->bracount, name, namelen,
4293                            (options & PCRE_EXTENDED) != 0)) > 0)
4294              {
4295              PUT2(code, 2+LINK_SIZE, i);
4296              }
4297    
4298            /* If terminator == 0 it means that the name followed directly after
4299            the opening parenthesis [e.g. (?(abc)...] and in this case there are
4300            some further alternatives to try. For the cases where terminator != 0
4301            [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
4302            now checked all the possibilities, so give an error. */
4303    
4304            else if (terminator != 0)
4305              {
4306              *errorcodeptr = ERR15;
4307              goto FAILED;
4308              }
4309    
4310            /* Check for (?(R) for recursion. Allow digits after R to specify a
4311            specific group number. */
4312    
4313            else if (*name == 'R')
4314              {
4315              recno = 0;
4316              for (i = 1; i < namelen; i++)
4317                {
4318                if ((digitab[name[i]] & ctype_digit) == 0)