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

Diff of /code/trunk/pcre_compile.c

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

revision 85 by nigel, Sat Feb 24 21:41:13 2007 UTC revision 264 by ph10, Tue Nov 13 11:07:16 2007 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2007 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
# Line 53  used by pcretest. DEBUG is not defined w Line 61  used by pcretest. DEBUG is not defined w
61  #endif  #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
79    
80  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
81  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
82  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
83  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
84  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
85    so this number is very generous.
86    
87    The same workspace is used during the second, actual compile phase for
88    remembering forward references to groups so that they can be filled in at the
89    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
90    is 4 there is plenty of room. */
91    
92  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
93    
94    
95  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 72  are simple data values; negative values Line 97  are simple data values; negative values
97  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
98  is invalid. */  is invalid. */
99    
100  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */
101  static const short int escapes[] = {  static const short int escapes[] = {
102       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */
103       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */
104     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */
105       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */
106  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */
107  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */
108     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */
109       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */
110  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */
111       0,      0, -ESC_z                                            /* x - z */       0,      0, -ESC_z                                            /* x - z */
112  };  };
113    
114  #else         /* This is the "abnormal" table for EBCDIC systems */  #else           /* This is the "abnormal" table for EBCDIC systems */
115  static const short int escapes[] = {  static const short int escapes[] = {
116  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
117  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 96  static const short int escapes[] = { Line 121  static const short int escapes[] = {
121  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
122  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
123  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
124  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
125  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
126  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
127  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
128  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
129  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
130  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
131  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
132  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
133  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
134  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
135  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
136  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
137  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
138  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 115  static const short int escapes[] = { Line 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
144  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
145  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. */
146    
147  static const char *const posix_names[] = {  typedef struct verbitem {
148    "alpha", "lower", "upper",    int   len;
149    "alnum", "ascii", "blank", "cntrl", "digit", "graph",    int   op;
150    "print", "punct", "space", "word",  "xdigit" };  } verbitem;
151    
152    static const char verbnames[] =
153      "ACCEPT\0"
154      "COMMIT\0"
155      "F\0"
156      "FAIL\0"
157      "PRUNE\0"
158      "SKIP\0"
159      "THEN";
160    
161    static verbitem verbs[] = {
162      { 6, OP_ACCEPT },
163      { 6, OP_COMMIT },
164      { 1, OP_FAIL },
165      { 4, OP_FAIL },
166      { 5, OP_PRUNE },
167      { 4, OP_SKIP  },
168      { 4, OP_THEN  }
169    };
170    
171    static int verbcount = sizeof(verbs)/sizeof(verbitem);
172    
173    
174    /* Tables of names of POSIX character classes and their lengths. The names are
175    now all in a single string, to reduce the number of relocations when a shared
176    library is dynamically loaded. The list of lengths is terminated by a zero
177    length entry. The first three must be alpha, lower, upper, as this is assumed
178    for handling case independence. */
179    
180    static const char posix_names[] =
181      "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"
182      "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"
183      "word\0"   "xdigit";
184    
185  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
186    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
187    
188  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
189  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
190  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
191    characters are removed, and for [:alpha:] and [:alnum:] the underscore
192    character is removed. The triples in the table consist of the base map offset,
193    second map offset or -1 if no second map, and a non-negative value for map
194    addition or a negative value for map subtraction (if there are two maps). The
195    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
196    remove vertical space characters, 2 => remove underscore. */
197    
198  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
199    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
200    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
201    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
202    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
203    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
204    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
205    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
206    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
207    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
208    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
209    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
210    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
211    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
212    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
213  };  };
214    
215    
216  /* The texts of compile-time error messages. These are "char *" because they  #define STRING(a)  # a
217  are passed to the outside world. */  #define XSTRING(s) STRING(s)
218    
219  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
220    "no error",  are passed to the outside world. Do not ever re-use any error number, because
221    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
222    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
223    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
224    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
225    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
226    simply count through to the one we want - this isn't a performance issue
227    because these strings are used only when there is a compilation error. */
228    
229    static const char error_texts[] =
230      "no error\0"
231      "\\ at end of pattern\0"
232      "\\c at end of pattern\0"
233      "unrecognized character follows \\\0"
234      "numbers out of order in {} quantifier\0"
235    /* 5 */    /* 5 */
236    "number too big in {} quantifier",    "number too big in {} quantifier\0"
237    "missing terminating ] for character class",    "missing terminating ] for character class\0"
238    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
239    "range out of order in character class",    "range out of order in character class\0"
240    "nothing to repeat",    "nothing to repeat\0"
241    /* 10 */    /* 10 */
242    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
243    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
244    "unrecognized character after (?",    "unrecognized character after (?\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";
304    
305    
306  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 229  For convenience, we use the same bit def Line 319  For convenience, we use the same bit def
319    
320  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
321    
322  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */
323  static const unsigned char digitab[] =  static const unsigned char digitab[] =
324    {    {
325    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 265  static const unsigned char digitab[] = Line 355  static const unsigned char digitab[] =
355    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
356    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
357    
358  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else           /* This is the "abnormal" case, for EBCDIC systems */
359  static const unsigned char digitab[] =  static const unsigned char digitab[] =
360    {    {
361    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 279  static const unsigned char digitab[] = Line 369  static const unsigned char digitab[] =
369    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
370    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
371    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
372    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
373    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
374    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
375    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 313  static const unsigned char ebcdic_charta Line 403  static const unsigned char ebcdic_charta
403    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
404    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
405    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
406    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
407    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
408    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
409    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 340  static const unsigned char ebcdic_charta Line 430  static const unsigned char ebcdic_charta
430  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
431    
432  static BOOL  static BOOL
433    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
434      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
435    
436    
437    
438    /*************************************************
439    *            Find an error text                  *
440    *************************************************/
441    
442    /* The error texts are now all in one long string, to save on relocations. As
443    some of the text is of unknown length, we can't use a table of offsets.
444    Instead, just count through the strings. This is not a performance issue
445    because it happens only when there has been a compilation error.
446    
447    Argument:   the error number
448    Returns:    pointer to the error string
449    */
450    
451    static const char *
452    find_error_text(int n)
453    {
454    const char *s = error_texts;
455    for (; n > 0; n--) while (*s++ != 0);
456    return s;
457    }
458    
459    
460  /*************************************************  /*************************************************
# Line 351  static BOOL Line 463  static BOOL
463    
464  /* 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
465  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
466  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
467  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
468  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,
469    ptr is pointing at the \. On exit, it is on the final character of the escape
470    sequence.
471    
472  Arguments:  Arguments:
473    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 364  Arguments: Line 478  Arguments:
478    
479  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
480                   negative => a special escape sequence                   negative => a special escape sequence
481                   on error, errorptr is set                   on error, errorcodeptr is set
482  */  */
483    
484  static int  static int
485  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
486    int options, BOOL isclass)    int options, BOOL isclass)
487  {  {
488  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
489    const uschar *ptr = *ptrptr + 1;
490  int c, i;  int c, i;
491    
492    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
493    ptr--;                            /* Set pointer back to the last byte */
494    
495  /* 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. */
496    
 c = *(++ptr);  
497  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
498    
499  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in
500  a table. A non-zero result is something that can be returned immediately.  a table. A non-zero result is something that can be returned immediately.
501  Otherwise further processing may be required. */  Otherwise further processing may be required. */
502    
503  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
504  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphameric */
505  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
506    
507  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
508  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */
509  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
510  #endif  #endif
# Line 397  else if ((i = escapes[c - 0x48]) != 0) Line 514  else if ((i = escapes[c - 0x48]) != 0)
514  else  else
515    {    {
516    const uschar *oldptr;    const uschar *oldptr;
517      BOOL braced, negated;
518    
519    switch (c)    switch (c)
520      {      {
521      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
# Line 410  else Line 529  else
529      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
530      break;      break;
531    
532        /* \g must be followed by a number, either plain or braced. If positive, it
533        is an absolute backreference. If negative, it is a relative backreference.
534        This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a
535        reference to a named group. This is part of Perl's movement towards a
536        unified syntax for back references. As this is synonymous with \k{name}, we
537        fudge it up by pretending it really was \k. */
538    
539        case 'g':
540        if (ptr[1] == '{')
541          {
542          const uschar *p;
543          for (p = ptr+2; *p != 0 && *p != '}'; p++)
544            if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;
545          if (*p != 0 && *p != '}')
546            {
547            c = -ESC_k;
548            break;
549            }
550          braced = TRUE;
551          ptr++;
552          }
553        else braced = FALSE;
554    
555        if (ptr[1] == '-')
556          {
557          negated = TRUE;
558          ptr++;
559          }
560        else negated = FALSE;
561    
562        c = 0;
563        while ((digitab[ptr[1]] & ctype_digit) != 0)
564          c = c * 10 + *(++ptr) - '0';
565    
566        if (c < 0)
567          {
568          *errorcodeptr = ERR61;
569          break;
570          }
571    
572        if (c == 0 || (braced && *(++ptr) != '}'))
573          {
574          *errorcodeptr = ERR57;
575          break;
576          }
577    
578        if (negated)
579          {
580          if (c > bracount)
581            {
582            *errorcodeptr = ERR15;
583            break;
584            }
585          c = bracount - (c - 1);
586          }
587    
588        c = -(ESC_REF + c);
589        break;
590    
591      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
592      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
593      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 431  else Line 609  else
609        c -= '0';        c -= '0';
610        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
611          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
612          if (c < 0)
613            {
614            *errorcodeptr = ERR61;
615            break;
616            }
617        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
618          {          {
619          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 451  else Line 634  else
634        }        }
635    
636      /* \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
637      larger first octal digit. */      larger first octal digit. The original code used just to take the least
638        significant 8 bits of octal numbers (I think this is what early Perls used
639        to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
640        than 3 octal digits. */
641    
642      case '0':      case '0':
643      c -= '0';      c -= '0';
644      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')
645          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - '0';
646      c &= 255;     /* Take least significant 8 bits */      if (!utf8 && c > 255) *errorcodeptr = ERR51;
647      break;      break;
648    
649      /* \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
650      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
651        treated as a data character. */
652    
653      case 'x':      case 'x':
654  #ifdef SUPPORT_UTF8      if (ptr[1] == '{')
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
655        {        {
656        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
657        register int count = 0;        int count = 0;
658    
659        c = 0;        c = 0;
660        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
661          {          {
662          int cc = *pt++;          register int cc = *pt++;
663            if (c == 0 && cc == '0') continue;     /* Leading zeroes */
664          count++;          count++;
665  #if !EBCDIC    /* ASCII coding */  
666    #ifndef EBCDIC  /* ASCII coding */
667          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= 'a') cc -= 32;               /* Convert to upper case */
668          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));
669  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
670          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */
671          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));
672  #endif  #endif
673          }          }
674    
675        if (*pt == '}')        if (*pt == '}')
676          {          {
677          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
678          ptr = pt;          ptr = pt;
679          break;          break;
680          }          }
681    
682        /* 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
683        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
684        }        }
 #endif  
685    
686      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
687    
688      c = 0;      c = 0;
689      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
690        {        {
691        int cc;                               /* Some compilers don't like ++ */        int cc;                               /* Some compilers don't like ++ */
692        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                        /* in initializers */
693  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
694        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= 'a') cc -= 32;              /* Convert to upper case */
695        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));
696  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
697        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= 'z') cc += 64;              /* Convert to upper case */
698        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));
699  #endif  #endif
700        }        }
701      break;      break;
702    
703      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
704        This coding is ASCII-specific, but then the whole concept of \cx is
705        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
706    
707      case 'c':      case 'c':
708      c = *(++ptr);      c = *(++ptr);
709      if (c == 0)      if (c == 0)
710        {        {
711        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
712        return 0;        break;
713        }        }
714    
715      /* 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 */  
716      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= 'a' && c <= 'z') c -= 32;
717      c ^= 0x40;      c ^= 0x40;
718  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
719      if (c >= 'a' && c <= 'z') c += 64;      if (c >= 'a' && c <= 'z') c += 64;
720      c ^= 0xC0;      c ^= 0xC0;
721  #endif  #endif
# Line 569  escape sequence. Line 757  escape sequence.
757  Argument:  Argument:
758    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
759    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
760      dptr           points to an int that is set to the detailed property value
761    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
762    
763  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
764  */  */
765    
766  static int  static int
767  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
768  {  {
769  int c, i, bot, top;  int c, i, bot, top;
770  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
771  char name[4];  char name[32];
772    
773  c = *(++ptr);  c = *(++ptr);
774  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
775    
776  *negptr = FALSE;  *negptr = FALSE;
777    
778  /* \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
779  preceded by ^ for negation. */  negation. */
780    
781  if (c == '{')  if (c == '{')
782    {    {
# Line 596  if (c == '{') Line 785  if (c == '{')
785      *negptr = TRUE;      *negptr = TRUE;
786      ptr++;      ptr++;
787      }      }
788    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
789      {      {
790      c = *(++ptr);      c = *(++ptr);
791      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
792      if (c == '}') break;      if (c == '}') break;
793      name[i] = c;      name[i] = c;
794      }      }
795    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;  
     }  
796    name[i] = 0;    name[i] = 0;
797    }    }
798    
# Line 628  top = _pcre_utt_size; Line 813  top = _pcre_utt_size;
813    
814  while (bot < top)  while (bot < top)
815    {    {
816    i = (bot + top)/2;    i = (bot + top) >> 1;
817    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
818    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
819        {
820        *dptr = _pcre_utt[i].value;
821        return _pcre_utt[i].type;
822        }
823    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
824    }    }
825    
 UNKNOWN_RETURN:  
826  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
827  *ptrptr = ptr;  *ptrptr = ptr;
828  return -1;  return -1;
# Line 750  return p; Line 938  return p;
938    
939    
940  /*************************************************  /*************************************************
941    *       Find forward referenced subpattern       *
942    *************************************************/
943    
944    /* This function scans along a pattern's text looking for capturing
945    subpatterns, and counting them. If it finds a named pattern that matches the
946    name it is given, it returns its number. Alternatively, if the name is NULL, it
947    returns when it reaches a given numbered subpattern. This is used for forward
948    references to subpatterns. We know that if (?P< is encountered, the name will
949    be terminated by '>' because that is checked in the first pass.
950    
951    Arguments:
952      ptr          current position in the pattern
953      count        current count of capturing parens so far encountered
954      name         name to seek, or NULL if seeking a numbered subpattern
955      lorn         name length, or subpattern number if name is NULL
956      xmode        TRUE if we are in /x mode
957    
958    Returns:       the number of the named subpattern, or -1 if not found
959    */
960    
961    static int
962    find_parens(const uschar *ptr, int count, const uschar *name, int lorn,
963      BOOL xmode)
964    {
965    const uschar *thisname;
966    
967    for (; *ptr != 0; ptr++)
968      {
969      int term;
970    
971      /* Skip over backslashed characters and also entire \Q...\E */
972    
973      if (*ptr == '\\')
974        {
975        if (*(++ptr) == 0) return -1;
976        if (*ptr == 'Q') for (;;)
977          {
978          while (*(++ptr) != 0 && *ptr != '\\');
979          if (*ptr == 0) return -1;
980          if (*(++ptr) == 'E') break;
981          }
982        continue;
983        }
984    
985      /* Skip over character classes */
986    
987      if (*ptr == '[')
988        {
989        while (*(++ptr) != ']')
990          {
991          if (*ptr == 0) return -1;
992          if (*ptr == '\\')
993            {
994            if (*(++ptr) == 0) return -1;
995            if (*ptr == 'Q') for (;;)
996              {
997              while (*(++ptr) != 0 && *ptr != '\\');
998              if (*ptr == 0) return -1;
999              if (*(++ptr) == 'E') break;
1000              }
1001            continue;
1002            }
1003          }
1004        continue;
1005        }
1006    
1007      /* Skip comments in /x mode */
1008    
1009      if (xmode && *ptr == '#')
1010        {
1011        while (*(++ptr) != 0 && *ptr != '\n');
1012        if (*ptr == 0) return -1;
1013        continue;
1014        }
1015    
1016      /* An opening parens must now be a real metacharacter */
1017    
1018      if (*ptr != '(') continue;
1019      if (ptr[1] != '?' && ptr[1] != '*')
1020        {
1021        count++;
1022        if (name == NULL && count == lorn) return count;
1023        continue;
1024        }
1025    
1026      ptr += 2;
1027      if (*ptr == 'P') ptr++;                      /* Allow optional P */
1028    
1029      /* We have to disambiguate (?<! and (?<= from (?<name> */
1030    
1031      if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&
1032           *ptr != '\'')
1033        continue;
1034    
1035      count++;
1036    
1037      if (name == NULL && count == lorn) return count;
1038      term = *ptr++;
1039      if (term == '<') term = '>';
1040      thisname = ptr;
1041      while (*ptr != term) ptr++;
1042      if (name != NULL && lorn == ptr - thisname &&
1043          strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1044        return count;
1045      }
1046    
1047    return -1;
1048    }
1049    
1050    
1051    
1052    /*************************************************
1053  *      Find first significant op code            *  *      Find first significant op code            *
1054  *************************************************/  *************************************************/
1055    
# Line 798  for (;;) Line 1098  for (;;)
1098    
1099      case OP_CALLOUT:      case OP_CALLOUT:
1100      case OP_CREF:      case OP_CREF:
1101      case OP_BRANUMBER:      case OP_RREF:
1102        case OP_DEF:
1103      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1104      break;      break;
1105    
# Line 843  for (;;) Line 1144  for (;;)
1144    {    {
1145    int d;    int d;
1146    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1147    switch (op)    switch (op)
1148      {      {
1149        case OP_CBRA:
1150      case OP_BRA:      case OP_BRA:
1151      case OP_ONCE:      case OP_ONCE:
1152      case OP_COND:      case OP_COND:
1153      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1154      if (d < 0) return d;      if (d < 0) return d;
1155      branchlength += d;      branchlength += d;
1156      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 885  for (;;) Line 1185  for (;;)
1185      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1186    
1187      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1188      case OP_CREF:      case OP_CREF:
1189        case OP_RREF:
1190        case OP_DEF:
1191      case OP_OPT:      case OP_OPT:
1192      case OP_CALLOUT:      case OP_CALLOUT:
1193      case OP_SOD:      case OP_SOD:
# Line 904  for (;;) Line 1205  for (;;)
1205    
1206      case OP_CHAR:      case OP_CHAR:
1207      case OP_CHARNC:      case OP_CHARNC:
1208        case OP_NOT:
1209      branchlength++;      branchlength++;
1210      cc += 2;      cc += 2;
1211  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 930  for (;;) Line 1232  for (;;)
1232    
1233      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1234      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1235        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1236      cc += 4;      cc += 4;
1237      break;      break;
1238    
# Line 937  for (;;) Line 1240  for (;;)
1240    
1241      case OP_PROP:      case OP_PROP:
1242      case OP_NOTPROP:      case OP_NOTPROP:
1243      cc++;      cc += 2;
1244      /* Fall through */      /* Fall through */
1245    
1246      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 1018  Returns: pointer to the opcode for Line 1321  Returns: pointer to the opcode for
1321  static const uschar *  static const uschar *
1322  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1323  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1324  for (;;)  for (;;)
1325    {    {
1326    register int c = *code;    register int c = *code;
1327    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1328    else if (c > OP_BRA)  
1329      /* XCLASS is used for classes that cannot be represented just by a bit
1330      map. This includes negated single high-valued characters. The length in
1331      the table is zero; the actual length is stored in the compiled code. */
1332    
1333      if (c == OP_XCLASS) code += GET(code, 1);
1334    
1335      /* Handle capturing bracket */
1336    
1337      else if (c == OP_CBRA)
1338      {      {
1339      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1340      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1341      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1342      }      }
1343    
1344      /* Otherwise, we can get the item's length from the table, except that for
1345      repeated character types, we have to test for \p and \P, which have an extra
1346      two bytes of parameters. */
1347    
1348    else    else
1349      {      {
1350      code += _pcre_OP_lengths[c];      switch(c)
1351          {
1352          case OP_TYPESTAR:
1353          case OP_TYPEMINSTAR:
1354          case OP_TYPEPLUS:
1355          case OP_TYPEMINPLUS:
1356          case OP_TYPEQUERY:
1357          case OP_TYPEMINQUERY:
1358          case OP_TYPEPOSSTAR:
1359          case OP_TYPEPOSPLUS:
1360          case OP_TYPEPOSQUERY:
1361          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1362          break;
1363    
1364  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1365          case OP_TYPEMINUPTO:
1366          case OP_TYPEEXACT:
1367          case OP_TYPEPOSUPTO:
1368          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1369          break;
1370          }
1371    
1372      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* Add in the fixed length from the table */
1373      by a multi-byte character. The length in the table is a minimum, so we have  
1374      to scan along to skip the extra bytes. All opcodes are less than 128, so we      code += _pcre_OP_lengths[c];
1375      can use relatively efficient code. */  
1376      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1377      a multi-byte character. The length in the table is a minimum, so we have to
1378      arrange to skip the extra bytes. */
1379    
1380    #ifdef SUPPORT_UTF8
1381      if (utf8) switch(c)      if (utf8) switch(c)
1382        {        {
1383        case OP_CHAR:        case OP_CHAR:
# Line 1051  for (;;) Line 1385  for (;;)
1385        case OP_EXACT:        case OP_EXACT:
1386        case OP_UPTO:        case OP_UPTO:
1387        case OP_MINUPTO:        case OP_MINUPTO:
1388          case OP_POSUPTO:
1389        case OP_STAR:        case OP_STAR:
1390        case OP_MINSTAR:        case OP_MINSTAR:
1391          case OP_POSSTAR:
1392        case OP_PLUS:        case OP_PLUS:
1393        case OP_MINPLUS:        case OP_MINPLUS:
1394          case OP_POSPLUS:
1395        case OP_QUERY:        case OP_QUERY:
1396        case OP_MINQUERY:        case OP_MINQUERY:
1397        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1398        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;  
1399        break;        break;
1400        }        }
1401  #endif  #endif
# Line 1092  Returns: pointer to the opcode for Line 1422  Returns: pointer to the opcode for
1422  static const uschar *  static const uschar *
1423  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1424  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1425  for (;;)  for (;;)
1426    {    {
1427    register int c = *code;    register int c = *code;
1428    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1429    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1430    else if (c > OP_BRA)  
1431      {    /* XCLASS is used for classes that cannot be represented just by a bit
1432      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1433      }    the table is zero; the actual length is stored in the compiled code. */
1434    
1435      if (c == OP_XCLASS) code += GET(code, 1);
1436    
1437      /* Otherwise, we can get the item's length from the table, except that for
1438      repeated character types, we have to test for \p and \P, which have an extra
1439      two bytes of parameters. */
1440    
1441    else    else
1442      {      {
1443      code += _pcre_OP_lengths[c];      switch(c)
1444          {
1445          case OP_TYPESTAR:
1446          case OP_TYPEMINSTAR:
1447          case OP_TYPEPLUS:
1448          case OP_TYPEMINPLUS:
1449          case OP_TYPEQUERY:
1450          case OP_TYPEMINQUERY:
1451          case OP_TYPEPOSSTAR:
1452          case OP_TYPEPOSPLUS:
1453          case OP_TYPEPOSQUERY:
1454          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1455          break;
1456    
1457  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1458          case OP_TYPEUPTO:
1459          case OP_TYPEMINUPTO:
1460          case OP_TYPEEXACT:
1461          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1462          break;
1463          }
1464    
1465        /* Add in the fixed length from the table */
1466    
1467        code += _pcre_OP_lengths[c];
1468    
1469      /* 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
1470      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
1471      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. */  
1472    
1473    #ifdef SUPPORT_UTF8
1474      if (utf8) switch(c)      if (utf8) switch(c)
1475        {        {
1476        case OP_CHAR:        case OP_CHAR:
# Line 1123  for (;;) Line 1478  for (;;)
1478        case OP_EXACT:        case OP_EXACT:
1479        case OP_UPTO:        case OP_UPTO:
1480        case OP_MINUPTO:        case OP_MINUPTO:
1481          case OP_POSUPTO:
1482        case OP_STAR:        case OP_STAR:
1483        case OP_MINSTAR:        case OP_MINSTAR:
1484          case OP_POSSTAR:
1485        case OP_PLUS:        case OP_PLUS:
1486        case OP_MINPLUS:        case OP_MINPLUS:
1487          case OP_POSPLUS:
1488        case OP_QUERY:        case OP_QUERY:
1489        case OP_MINQUERY:        case OP_MINQUERY:
1490        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1491        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;  
1492        break;        break;
1493        }        }
1494  #endif  #endif
# Line 1152  for (;;) Line 1503  for (;;)
1503  *************************************************/  *************************************************/
1504    
1505  /* 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
1506  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()
1507  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
1508  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
1509  whose current branch will already have been scanned.  assertions. If we hit an unclosed bracket, we return "empty" - this means we've
1510    struck an inner bracket whose current branch will already have been scanned.
1511    
1512  Arguments:  Arguments:
1513    code        points to start of search    code        points to start of search
# Line 1169  static BOOL Line 1521  static BOOL
1521  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1522  {  {
1523  register int c;  register int c;
1524  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);
1525       code < endcode;       code < endcode;
1526       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1527    {    {
# Line 1177  for (code = first_significant_code(code Line 1529  for (code = first_significant_code(code
1529    
1530    c = *code;    c = *code;
1531    
1532    if (c >= OP_BRA)    /* Groups with zero repeats can of course be empty; skip them. */
1533    
1534      if (c == OP_BRAZERO || c == OP_BRAMINZERO)
1535        {
1536        code += _pcre_OP_lengths[c];
1537        do code += GET(code, 1); while (*code == OP_ALT);
1538        c = *code;
1539        continue;
1540        }
1541    
1542      /* For other groups, scan the branches. */
1543    
1544      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1545      {      {
1546      BOOL empty_branch;      BOOL empty_branch;
1547      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1193  for (code = first_significant_code(code Line 1557  for (code = first_significant_code(code
1557        }        }
1558      while (*code == OP_ALT);      while (*code == OP_ALT);
1559      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
     code += 1 + LINK_SIZE;  
1560      c = *code;      c = *code;
1561        continue;
1562      }      }
1563    
1564    else switch (c)    /* Handle the other opcodes */
1565    
1566      switch (c)
1567      {      {
1568      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1569        cannot be represented just by a bit map. This includes negated single
1570        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1571        actual length is stored in the compiled code, so we must update "code"
1572        here. */
1573    
1574  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1575      case OP_XCLASS:      case OP_XCLASS:
1576      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1577      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1578  #endif  #endif
1579    
# Line 1253  for (code = first_significant_code(code Line 1623  for (code = first_significant_code(code
1623      case OP_NOT:      case OP_NOT:
1624      case OP_PLUS:      case OP_PLUS:
1625      case OP_MINPLUS:      case OP_MINPLUS:
1626        case OP_POSPLUS:
1627      case OP_EXACT:      case OP_EXACT:
1628      case OP_NOTPLUS:      case OP_NOTPLUS:
1629      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1630        case OP_NOTPOSPLUS:
1631      case OP_NOTEXACT:      case OP_NOTEXACT:
1632      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1633      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1634        case OP_TYPEPOSPLUS:
1635      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1636      return FALSE;      return FALSE;
1637    
1638        /* These are going to continue, as they may be empty, but we have to
1639        fudge the length for the \p and \P cases. */
1640    
1641        case OP_TYPESTAR:
1642        case OP_TYPEMINSTAR:
1643        case OP_TYPEPOSSTAR:
1644        case OP_TYPEQUERY:
1645        case OP_TYPEMINQUERY:
1646        case OP_TYPEPOSQUERY:
1647        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1648        break;
1649    
1650        /* Same for these */
1651    
1652        case OP_TYPEUPTO:
1653        case OP_TYPEMINUPTO:
1654        case OP_TYPEPOSUPTO:
1655        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1656        break;
1657    
1658      /* End of branch */      /* End of branch */
1659    
1660      case OP_KET:      case OP_KET:
# Line 1270  for (code = first_significant_code(code Line 1663  for (code = first_significant_code(code
1663      case OP_ALT:      case OP_ALT:
1664      return TRUE;      return TRUE;
1665    
1666      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1667      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1668    
1669  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1670      case OP_STAR:      case OP_STAR:
1671      case OP_MINSTAR:      case OP_MINSTAR:
1672        case OP_POSSTAR:
1673      case OP_QUERY:      case OP_QUERY:
1674      case OP_MINQUERY:      case OP_MINQUERY:
1675        case OP_POSQUERY:
1676      case OP_UPTO:      case OP_UPTO:
1677      case OP_MINUPTO:      case OP_MINUPTO:
1678        case OP_POSUPTO:
1679      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1680      break;      break;
1681  #endif  #endif
# Line 1375  Returns: a value representing the na Line 1771  Returns: a value representing the na
1771  static int  static int
1772  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1773  {  {
1774    const char *pn = posix_names;
1775  register int yield = 0;  register int yield = 0;
1776  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1777    {    {
1778    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1779      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1780      pn += posix_name_lengths[yield] + 1;
1781    yield++;    yield++;
1782    }    }
1783  return -1;  return -1;
# Line 1397  earlier groups that are outside the curr Line 1795  earlier groups that are outside the curr
1795  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
1796  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
1797  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
1798  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,
1799  partially compiled regex must be temporarily terminated with OP_END.  the partially compiled regex must be temporarily terminated with OP_END.
1800    
1801    This function has been extended with the possibility of forward references for
1802    recursions and subroutine calls. It must also check the list of such references
1803    for the group we are dealing with. If it finds that one of the recursions in
1804    the current group is on this list, it adjusts the offset in the list, not the
1805    value in the reference (which is a group number).
1806    
1807  Arguments:  Arguments:
1808    group      points to the start of the group    group      points to the start of the group
1809    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1810    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1811    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1812      save_hwm   the hwm forward reference pointer at the start of the group
1813    
1814  Returns:     nothing  Returns:     nothing
1815  */  */
1816    
1817  static void  static void
1818  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1819      uschar *save_hwm)
1820  {  {
1821  uschar *ptr = group;  uschar *ptr = group;
1822    
1823  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1824    {    {
1825    int offset = GET(ptr, 1);    int offset;
1826    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1827    
1828      /* See if this recursion is on the forward reference list. If so, adjust the
1829      reference. */
1830    
1831      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1832        {
1833        offset = GET(hc, 0);
1834        if (cd->start_code + offset == ptr + 1)
1835          {
1836          PUT(hc, 0, offset + adjust);
1837          break;
1838          }
1839        }
1840    
1841      /* Otherwise, adjust the recursion offset if it's after the start of this
1842      group. */
1843    
1844      if (hc >= cd->hwm)
1845        {
1846        offset = GET(ptr, 1);
1847        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1848        }
1849    
1850    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1851    }    }
1852  }  }
# Line 1495  Yield: TRUE when range returned; Line 1925  Yield: TRUE when range returned;
1925  */  */
1926    
1927  static BOOL  static BOOL
1928  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
1929      unsigned int *odptr)
1930  {  {
1931  int c, chartype, othercase, next;  unsigned int c, othercase, next;
1932    
1933  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
1934    {    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
1935    
1936  if (c > d) return FALSE;  if (c > d) return FALSE;
1937    
# Line 1512  next = othercase + 1; Line 1940  next = othercase + 1;
1940    
1941  for (++c; c <= d; c++)  for (++c; c <= d; c++)
1942    {    {
1943    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (_pcre_ucp_othercase(c) != next) break;
         othercase != next)  
     break;  
1944    next++;    next++;
1945    }    }
1946    
# Line 1526  return TRUE; Line 1952  return TRUE;
1952  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
1953    
1954    
1955    
1956  /*************************************************  /*************************************************
1957  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
1958  *************************************************/  *************************************************/
1959    
1960  /* 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
1961  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
1962  bits.  sense to automatically possessify the repeated item.
1963    
1964  Arguments:  Arguments:
1965    optionsptr     pointer to the option bits    op_code       the repeated op code
1966    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
1967    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
1968    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
1969    errorcodeptr   points to error code variable    ptr           next character in pattern
1970    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
1971    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.  
1972    
1973  Returns:         TRUE on success  Returns:        TRUE if possessifying is wanted
                  FALSE, with *errorcodeptr set non-zero on error  
1974  */  */
1975    
1976  static BOOL  static BOOL
1977  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
1978    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const uschar *ptr, int options, compile_data *cd)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
1979  {  {
1980  int repeat_type, op_type;  int next;
 int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  
 int bravalue = 0;  
 int greedy_default, greedy_non_default;  
 int firstbyte, reqbyte;  
 int zeroreqbyte, zerofirstbyte;  
 int req_caseopt, reqvary, tempreqvary;  
 int condcount = 0;  
 int options = *optionsptr;  
 int after_manual_callout = 0;  
 register int c;  
 register uschar *code = *codeptr;  
 uschar *tempcode;  
 BOOL inescq = FALSE;  
 BOOL groupsetfirstbyte = FALSE;  
 const uschar *ptr = *ptrptr;  
 const uschar *tempptr;  
 uschar *previous = NULL;  
 uschar *previous_callout = NULL;  
 uschar classbits[32];  
   
 #ifdef SUPPORT_UTF8  
 BOOL class_utf8;  
 BOOL utf8 = (options & PCRE_UTF8) != 0;  
 uschar *class_utf8data;  
 uschar utf8_char[6];  
 #else  
 BOOL utf8 = FALSE;  
 #endif  
1981    
1982  /* Set up the default and non-default settings for greediness */  /* Skip whitespace and comments in extended mode */
1983    
1984    if ((options & PCRE_EXTENDED) != 0)
1985      {
1986      for (;;)
1987        {
1988        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
1989        if (*ptr == '#')
1990          {
1991          while (*(++ptr) != 0)
1992            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
1993          }
1994        else break;
1995        }
1996      }
1997    
1998    /* If the next item is one that we can handle, get its value. A non-negative
1999    value is a character, a negative value is an escape value. */
2000    
2001    if (*ptr == '\\')
2002      {
2003      int temperrorcode = 0;
2004      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2005      if (temperrorcode != 0) return FALSE;
2006      ptr++;    /* Point after the escape sequence */
2007      }
2008    
2009    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2010      {
2011    #ifdef SUPPORT_UTF8
2012      if (utf8) { GETCHARINC(next, ptr); } else
2013    #endif
2014      next = *ptr++;
2015      }
2016    
2017    else return FALSE;
2018    
2019    /* Skip whitespace and comments in extended mode */
2020    
2021    if ((options & PCRE_EXTENDED) != 0)
2022      {
2023      for (;;)
2024        {
2025        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2026        if (*ptr == '#')
2027          {
2028          while (*(++ptr) != 0)
2029            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2030          }
2031        else break;
2032        }
2033      }
2034    
2035    /* If the next thing is itself optional, we have to give up. */
2036    
2037    if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
2038      return FALSE;
2039    
2040    /* Now compare the next item with the previous opcode. If the previous is a
2041    positive single character match, "item" either contains the character or, if
2042    "item" is greater than 127 in utf8 mode, the character's bytes are in
2043    utf8_char. */
2044    
2045    
2046    /* Handle cases when the next item is a character. */
2047    
2048    if (next >= 0) switch(op_code)
2049      {
2050      case OP_CHAR:
2051    #ifdef SUPPORT_UTF8
2052      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2053    #endif
2054      return item != next;
2055    
2056      /* For CHARNC (caseless character) we must check the other case. If we have
2057      Unicode property support, we can use it to test the other case of
2058      high-valued characters. */
2059    
2060      case OP_CHARNC:
2061    #ifdef SUPPORT_UTF8
2062      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2063    #endif
2064      if (item == next) return FALSE;
2065    #ifdef SUPPORT_UTF8
2066      if (utf8)
2067        {
2068        unsigned int othercase;
2069        if (next < 128) othercase = cd->fcc[next]; else
2070    #ifdef SUPPORT_UCP
2071        othercase = _pcre_ucp_othercase((unsigned int)next);
2072    #else
2073        othercase = NOTACHAR;
2074    #endif
2075        return (unsigned int)item != othercase;
2076        }
2077      else
2078    #endif  /* SUPPORT_UTF8 */
2079      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2080    
2081      /* For OP_NOT, "item" must be a single-byte character. */
2082    
2083      case OP_NOT:
2084      if (next < 0) return FALSE;  /* Not a character */
2085      if (item == next) return TRUE;
2086      if ((options & PCRE_CASELESS) == 0) return FALSE;
2087    #ifdef SUPPORT_UTF8
2088      if (utf8)
2089        {
2090        unsigned int othercase;
2091        if (next < 128) othercase = cd->fcc[next]; else
2092    #ifdef SUPPORT_UCP
2093        othercase = _pcre_ucp_othercase(next);
2094    #else
2095        othercase = NOTACHAR;
2096    #endif
2097        return (unsigned int)item == othercase;
2098        }
2099      else
2100    #endif  /* SUPPORT_UTF8 */
2101      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2102    
2103      case OP_DIGIT:
2104      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2105    
2106      case OP_NOT_DIGIT:
2107      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2108    
2109      case OP_WHITESPACE:
2110      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2111    
2112      case OP_NOT_WHITESPACE:
2113      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2114    
2115      case OP_WORDCHAR:
2116      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2117    
2118      case OP_NOT_WORDCHAR:
2119      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2120    
2121      case OP_HSPACE:
2122      case OP_NOT_HSPACE:
2123      switch(next)
2124        {
2125        case 0x09:
2126        case 0x20:
2127        case 0xa0:
2128        case 0x1680:
2129        case 0x180e:
2130        case 0x2000:
2131        case 0x2001:
2132        case 0x2002:
2133        case 0x2003:
2134        case 0x2004:
2135        case 0x2005:
2136        case 0x2006:
2137        case 0x2007:
2138        case 0x2008:
2139        case 0x2009:
2140        case 0x200A:
2141        case 0x202f:
2142        case 0x205f:
2143        case 0x3000:
2144        return op_code != OP_HSPACE;
2145        default:
2146        return op_code == OP_HSPACE;
2147        }
2148    
2149      case OP_VSPACE:
2150      case OP_NOT_VSPACE:
2151      switch(next)
2152        {
2153        case 0x0a:
2154        case 0x0b:
2155        case 0x0c:
2156        case 0x0d:
2157        case 0x85:
2158        case 0x2028:
2159        case 0x2029:
2160        return op_code != OP_VSPACE;
2161        default:
2162        return op_code == OP_VSPACE;
2163        }
2164    
2165      default:
2166      return FALSE;
2167      }
2168    
2169    
2170    /* Handle the case when the next item is \d, \s, etc. */
2171    
2172    switch(op_code)
2173      {
2174      case OP_CHAR:
2175      case OP_CHARNC:
2176    #ifdef SUPPORT_UTF8
2177      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2178    #endif
2179      switch(-next)
2180        {
2181        case ESC_d:
2182        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2183    
2184        case ESC_D:
2185        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2186    
2187        case ESC_s:
2188        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2189    
2190        case ESC_S:
2191        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2192    
2193        case ESC_w:
2194        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2195    
2196        case ESC_W:
2197        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2198    
2199        case ESC_h:
2200        case ESC_H:
2201        switch(item)
2202          {
2203          case 0x09:
2204          case 0x20:
2205          case 0xa0:
2206          case 0x1680:
2207          case 0x180e:
2208          case 0x2000:
2209          case 0x2001:
2210          case 0x2002:
2211          case 0x2003:
2212          case 0x2004:
2213          case 0x2005:
2214          case 0x2006:
2215          case 0x2007:
2216          case 0x2008:
2217          case 0x2009:
2218          case 0x200A:
2219          case 0x202f:
2220          case 0x205f:
2221          case 0x3000:
2222          return -next != ESC_h;
2223          default:
2224          return -next == ESC_h;
2225          }
2226    
2227        case ESC_v:
2228        case ESC_V:
2229        switch(item)
2230          {
2231          case 0x0a:
2232          case 0x0b:
2233          case 0x0c:
2234          case 0x0d:
2235          case 0x85:
2236          case 0x2028:
2237          case 0x2029:
2238          return -next != ESC_v;
2239          default:
2240          return -next == ESC_v;
2241          }
2242    
2243        default:
2244        return FALSE;
2245        }
2246    
2247      case OP_DIGIT:
2248      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2249             next == -ESC_h || next == -ESC_v;
2250    
2251      case OP_NOT_DIGIT:
2252      return next == -ESC_d;
2253    
2254      case OP_WHITESPACE:
2255      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2256    
2257      case OP_NOT_WHITESPACE:
2258      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2259    
2260      case OP_HSPACE:
2261      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2262    
2263      case OP_NOT_HSPACE:
2264      return next == -ESC_h;
2265    
2266      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2267      case OP_VSPACE:
2268      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2269    
2270      case OP_NOT_VSPACE:
2271      return next == -ESC_v;
2272    
2273      case OP_WORDCHAR:
2274      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2275    
2276      case OP_NOT_WORDCHAR:
2277      return next == -ESC_w || next == -ESC_d;
2278    
2279      default:
2280      return FALSE;
2281      }
2282    
2283    /* Control does not reach here */
2284    }
2285    
2286    
2287    
2288    /*************************************************
2289    *           Compile one branch                   *
2290    *************************************************/
2291    
2292    /* Scan the pattern, compiling it into the a vector. If the options are
2293    changed during the branch, the pointer is used to change the external options
2294    bits. This function is used during the pre-compile phase when we are trying
2295    to find out the amount of memory needed, as well as during the real compile
2296    phase. The value of lengthptr distinguishes the two phases.
2297    
2298    Arguments:
2299      optionsptr     pointer to the option bits
2300      codeptr        points to the pointer to the current code point
2301      ptrptr         points to the current pattern pointer
2302      errorcodeptr   points to error code variable
2303      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2304      reqbyteptr     set to the last literal character required, else < 0
2305      bcptr          points to current branch chain
2306      cd             contains pointers to tables etc.
2307      lengthptr      NULL during the real compile phase
2308                     points to length accumulator during pre-compile phase
2309    
2310    Returns:         TRUE on success
2311                     FALSE, with *errorcodeptr set non-zero on error
2312    */
2313    
2314    static BOOL
2315    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2316      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2317      compile_data *cd, int *lengthptr)
2318    {
2319    int repeat_type, op_type;
2320    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2321    int bravalue = 0;
2322    int greedy_default, greedy_non_default;
2323    int firstbyte, reqbyte;
2324    int zeroreqbyte, zerofirstbyte;
2325    int req_caseopt, reqvary, tempreqvary;
2326    int options = *optionsptr;
2327    int after_manual_callout = 0;
2328    int length_prevgroup = 0;
2329    register int c;
2330    register uschar *code = *codeptr;
2331    uschar *last_code = code;
2332    uschar *orig_code = code;
2333    uschar *tempcode;
2334    BOOL inescq = FALSE;
2335    BOOL groupsetfirstbyte = FALSE;
2336    const uschar *ptr = *ptrptr;
2337    const uschar *tempptr;
2338    uschar *previous = NULL;
2339    uschar *previous_callout = NULL;
2340    uschar *save_hwm = NULL;
2341    uschar classbits[32];
2342    
2343    #ifdef SUPPORT_UTF8
2344    BOOL class_utf8;
2345    BOOL utf8 = (options & PCRE_UTF8) != 0;
2346    uschar *class_utf8data;
2347    uschar utf8_char[6];
2348    #else
2349    BOOL utf8 = FALSE;
2350    uschar *utf8_char = NULL;
2351    #endif
2352    
2353    #ifdef DEBUG
2354    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2355    #endif
2356    
2357    /* Set up the default and non-default settings for greediness */
2358    
2359  greedy_default = ((options & PCRE_UNGREEDY) != 0);  greedy_default = ((options & PCRE_UNGREEDY) != 0);
2360  greedy_non_default = greedy_default ^ 1;  greedy_non_default = greedy_default ^ 1;
# Line 1613  req_caseopt = ((options & PCRE_CASELESS) Line 2383  req_caseopt = ((options & PCRE_CASELESS)
2383  for (;; ptr++)  for (;; ptr++)
2384    {    {
2385    BOOL negate_class;    BOOL negate_class;
2386      BOOL should_flip_negation;
2387    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2388    BOOL is_quantifier;    BOOL is_quantifier;
2389      BOOL is_recurse;
2390      BOOL reset_bracount;
2391    int class_charcount;    int class_charcount;
2392    int class_lastchar;    int class_lastchar;
2393    int newoptions;    int newoptions;
2394    int recno;    int recno;
2395      int refsign;
2396    int skipbytes;    int skipbytes;
2397    int subreqbyte;    int subreqbyte;
2398    int subfirstbyte;    int subfirstbyte;
2399      int terminator;
2400    int mclength;    int mclength;
2401    uschar mcbuffer[8];    uschar mcbuffer[8];
2402    
2403    /* Next byte in the pattern */    /* Get next byte in the pattern */
2404    
2405    c = *ptr;    c = *ptr;
2406    
2407      /* If we are in the pre-compile phase, accumulate the length used for the
2408      previous cycle of this loop. */
2409    
2410      if (lengthptr != NULL)
2411        {
2412    #ifdef DEBUG
2413        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2414    #endif
2415        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2416          {
2417          *errorcodeptr = ERR52;
2418          goto FAILED;
2419          }
2420    
2421        /* There is at least one situation where code goes backwards: this is the
2422        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2423        the class is simply eliminated. However, it is created first, so we have to
2424        allow memory for it. Therefore, don't ever reduce the length at this point.
2425        */
2426    
2427        if (code < last_code) code = last_code;
2428    
2429        /* Paranoid check for integer overflow */
2430    
2431        if (OFLOW_MAX - *lengthptr < code - last_code)
2432          {
2433          *errorcodeptr = ERR20;
2434          goto FAILED;
2435          }
2436    
2437        *lengthptr += code - last_code;
2438        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2439    
2440        /* If "previous" is set and it is not at the start of the work space, move
2441        it back to there, in order to avoid filling up the work space. Otherwise,
2442        if "previous" is NULL, reset the current code pointer to the start. */
2443    
2444        if (previous != NULL)
2445          {
2446          if (previous > orig_code)
2447            {
2448            memmove(orig_code, previous, code - previous);
2449            code -= previous - orig_code;
2450            previous = orig_code;
2451            }
2452          }
2453        else code = orig_code;
2454    
2455        /* Remember where this code item starts so we can pick up the length
2456        next time round. */
2457    
2458        last_code = code;
2459        }
2460    
2461      /* In the real compile phase, just check the workspace used by the forward
2462      reference list. */
2463    
2464      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2465        {
2466        *errorcodeptr = ERR52;
2467        goto FAILED;
2468        }
2469    
2470    /* 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 */
2471    
2472    if (inescq && c != 0)    if (inescq && c != 0)
# Line 1643  for (;; ptr++) Line 2481  for (;; ptr++)
2481        {        {
2482        if (previous_callout != NULL)        if (previous_callout != NULL)
2483          {          {
2484          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2485              complete_callout(previous_callout, ptr, cd);
2486          previous_callout = NULL;          previous_callout = NULL;
2487          }          }
2488        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1664  for (;; ptr++) Line 2503  for (;; ptr++)
2503    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2504         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2505      {      {
2506      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2507          complete_callout(previous_callout, ptr, cd);
2508      previous_callout = NULL;      previous_callout = NULL;
2509      }      }
2510    
# Line 1675  for (;; ptr++) Line 2515  for (;; ptr++)
2515      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2516      if (c == '#')      if (c == '#')
2517        {        {
2518        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2519        on the Macintosh. */          {
2520        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2521        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2522          if (*ptr != 0) continue;
2523    
2524          /* Else fall through to handle end of string */
2525          c = 0;
2526        }        }
2527      }      }
2528    
# Line 1692  for (;; ptr++) Line 2536  for (;; ptr++)
2536    
2537    switch(c)    switch(c)
2538      {      {
2539      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2540        case 0:                        /* The branch terminates at string end */
2541      case 0:      case '|':                      /* or | or ) */
     case '|':  
2542      case ')':      case ')':
2543      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2544      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2545      *codeptr = code;      *codeptr = code;
2546      *ptrptr = ptr;      *ptrptr = ptr;
2547        if (lengthptr != NULL)
2548          {
2549          if (OFLOW_MAX - *lengthptr < code - last_code)
2550            {
2551            *errorcodeptr = ERR20;
2552            goto FAILED;
2553            }
2554          *lengthptr += code - last_code;   /* To include callout length */
2555          DPRINTF((">> end branch\n"));
2556          }
2557      return TRUE;      return TRUE;
2558    
2559    
2560        /* ===================================================================*/
2561      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2562      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2563    
# Line 1731  for (;; ptr++) Line 2586  for (;; ptr++)
2586      *code++ = OP_ANY;      *code++ = OP_ANY;
2587      break;      break;
2588    
2589      /* Character classes. If the included characters are all < 255 in value, we  
2590      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2591      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
2592      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
2593      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2594        map as usual, then invert it at the end. However, we use a different opcode
2595        so that data characters > 255 can be handled correctly.
2596    
2597      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2598      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
# Line 1756  for (;; ptr++) Line 2613  for (;; ptr++)
2613        goto FAILED;        goto FAILED;
2614        }        }
2615    
2616      /* 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,
2617        if the first few characters (either before or after ^) are \Q\E or \E we
2618        skip them too. This makes for compatibility with Perl. */
2619    
2620      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2621        for (;;)
2622        {        {
       negate_class = TRUE;  
2623        c = *(++ptr);        c = *(++ptr);
2624          if (c == '\\')
2625            {
2626            if (ptr[1] == 'E') ptr++;
2627              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2628                else break;
2629            }
2630          else if (!negate_class && c == '^')
2631            negate_class = TRUE;
2632          else break;
2633        }        }
2634      else  
2635        {      /* If a class contains a negative special such as \S, we need to flip the
2636        negate_class = FALSE;      negation flag at the end, so that support for characters > 255 works
2637        }      correctly (they are all included in the class). */
2638    
2639        should_flip_negation = FALSE;
2640    
2641      /* 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
2642      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
2643      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2644    
2645      class_charcount = 0;      class_charcount = 0;
2646      class_lastchar = -1;      class_lastchar = -1;
2647    
2648        /* Initialize the 32-char bit map to all zeros. We build the map in a
2649        temporary bit of memory, in case the class contains only 1 character (less
2650        than 256), because in that case the compiled code doesn't use the bit map.
2651        */
2652    
2653        memset(classbits, 0, 32 * sizeof(uschar));
2654    
2655  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2656      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2657      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2658  #endif  #endif
2659    
     /* 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));  
   
2660      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2661      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
2662      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. */  
2663    
2664      do      if (c != 0) do
2665        {        {
2666          const uschar *oldptr;
2667    
2668  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2669        if (utf8 && c > 127)        if (utf8 && c > 127)
2670          {                           /* Braces are required because the */          {                           /* Braces are required because the */
# Line 1806  for (;; ptr++) Line 2676  for (;; ptr++)
2676    
2677        if (inescq)        if (inescq)
2678          {          {
2679          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2680            {            {
2681            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2682            ptr++;            ptr++;                            /* Skip the 'E' */
2683            continue;            continue;                         /* Carry on with next */
2684            }            }
2685          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2686          }          }
2687    
2688        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1826  for (;; ptr++) Line 2696  for (;; ptr++)
2696            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr, cd))
2697          {          {
2698          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2699          int posix_class, i;          int posix_class, taboffset, tabopt;
2700          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2701            uschar pbits[32];
2702    
2703          if (ptr[1] != ':')          if (ptr[1] != ':')
2704            {            {
# Line 1856  for (;; ptr++) Line 2727  for (;; ptr++)
2727          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
2728            posix_class = 0;            posix_class = 0;
2729    
2730          /* 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
2731          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
2732          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
2733          white space chars afterwards. */          result into the bit map that is being built. */
2734    
2735          posix_class *= 3;          posix_class *= 3;
2736          for (i = 0; i < 3; i++)  
2737            /* Copy in the first table (always present) */
2738    
2739            memcpy(pbits, cbits + posix_class_maps[posix_class],
2740              32 * sizeof(uschar));
2741    
2742            /* If there is a second table, add or remove it as required. */
2743    
2744            taboffset = posix_class_maps[posix_class + 1];
2745            tabopt = posix_class_maps[posix_class + 2];
2746    
2747            if (taboffset >= 0)
2748            {            {
2749            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
2750            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;  
             }  
2751            else            else
2752              {              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;  
             }  
2753            }            }
2754    
2755            /* Not see if we need to remove any special characters. An option
2756            value of 1 removes vertical space and 2 removes underscore. */
2757    
2758            if (tabopt < 0) tabopt = -tabopt;
2759            if (tabopt == 1) pbits[1] &= ~0x3c;
2760              else if (tabopt == 2) pbits[11] &= 0x7f;
2761    
2762            /* Add the POSIX table or its complement into the main table that is
2763            being built and we are done. */
2764    
2765            if (local_negate)
2766              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
2767            else
2768              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
2769    
2770          ptr = tempptr + 1;          ptr = tempptr + 1;
2771          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
2772          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
2773          }          }
2774    
2775        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2776        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
2777        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.
2778        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2779        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  
2780        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2781    
2782        if (c == '\\')        if (c == '\\')
2783          {          {
2784          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2785            if (*errorcodeptr != 0) goto FAILED;
2786    
2787          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */
2788          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 */
2789            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2790          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2791            {            {
2792            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1910  for (;; ptr++) Line 2796  for (;; ptr++)
2796            else inescq = TRUE;            else inescq = TRUE;
2797            continue;            continue;
2798            }            }
2799            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2800    
2801          if (c < 0)          if (c < 0)
2802            {            {
2803            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2804            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2805            switch (-c)  
2806              /* Save time by not doing this in the pre-compile phase. */
2807    
2808              if (lengthptr == NULL) switch (-c)
2809              {              {
2810              case ESC_d:              case ESC_d:
2811              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
2812              continue;              continue;
2813    
2814              case ESC_D:              case ESC_D:
2815                should_flip_negation = TRUE;
2816              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2817              continue;              continue;
2818    
# Line 1930  for (;; ptr++) Line 2821  for (;; ptr++)
2821              continue;              continue;
2822    
2823              case ESC_W:              case ESC_W:
2824                should_flip_negation = TRUE;
2825              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2826              continue;              continue;
2827    
# Line 1939  for (;; ptr++) Line 2831  for (;; ptr++)
2831              continue;              continue;
2832    
2833              case ESC_S:              case ESC_S:
2834                should_flip_negation = TRUE;
2835              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2836              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2837              continue;              continue;
2838    
2839  #ifdef SUPPORT_UCP              case ESC_E: /* Perl ignores an orphan \E */
             case ESC_p:  
             case ESC_P:  
               {  
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
               class_utf8 = TRUE;  
               *class_utf8data++ = ((-c == ESC_p) != negated)?  
                 XCL_PROP : XCL_NOTPROP;  
               *class_utf8data++ = property;  
               class_charcount -= 2;   /* Not a < 256 character */  
               }  
2840              continue;              continue;
 #endif  
2841    
2842              /* Unrecognized escapes are faulted if PCRE is running in its              default:    /* Not recognized; fall through */
2843              strict mode. By default, for compatibility with Perl, they are              break;      /* Need "default" setting to stop compiler warning. */
             treated as literals. */  
   
             default:  
             if ((options & PCRE_EXTRA) != 0)  
               {  
               *errorcodeptr = ERR7;  
               goto FAILED;  
               }  
             c = *ptr;              /* The final character */  
             class_charcount -= 2;  /* Undo the default count from above */  
2844              }              }
           }  
2845    
2846          /* Fall through if we have a single character (c >= 0). This may be            /* In the pre-compile phase, just do the recognition. */
         > 256 in UTF-8 mode. */  
2847    
2848          }   /* End of backslash handling */            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2849                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
2850    
2851        /* A single character may be followed by '-' to form a range. However,            /* We need to deal with \H, \h, \V, and \v in both phases because
2852        Perl does not permit ']' to be the end of the range. A '-' character            they use extra memory. */
       here is treated as a literal. */  
2853    
2854        if (ptr[1] == '-' && ptr[2] != ']')            if (-c == ESC_h)
2855          {              {
2856          int d;              SETBIT(classbits, 0x09); /* VT */
2857          ptr += 2;              SETBIT(classbits, 0x20); /* SPACE */
2858                SETBIT(classbits, 0xa0); /* NSBP */
2859    #ifdef SUPPORT_UTF8
2860                if (utf8)
2861                  {
2862                  class_utf8 = TRUE;
2863                  *class_utf8data++ = XCL_SINGLE;
2864                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2865                  *class_utf8data++ = XCL_SINGLE;
2866                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2867                  *class_utf8data++ = XCL_RANGE;
2868                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2869                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2870                  *class_utf8data++ = XCL_SINGLE;
2871                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2872                  *class_utf8data++ = XCL_SINGLE;
2873                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2874                  *class_utf8data++ = XCL_SINGLE;
2875                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2876                  }
2877    #endif
2878                continue;
2879                }
2880    
2881              if (-c == ESC_H)
2882                {
2883                for (c = 0; c < 32; c++)
2884                  {
2885                  int x = 0xff;
2886                  switch (c)
2887                    {
2888                    case 0x09/8: x ^= 1 << (0x09%8); break;
2889                    case 0x20/8: x ^= 1 << (0x20%8); break;
2890                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2891                    default: break;
2892                    }
2893                  classbits[c] |= x;
2894                  }
2895    
2896    #ifdef SUPPORT_UTF8
2897                if (utf8)
2898                  {
2899                  class_utf8 = TRUE;
2900                  *class_utf8data++ = XCL_RANGE;
2901                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2902                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
2903                  *class_utf8data++ = XCL_RANGE;
2904                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
2905                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
2906                  *class_utf8data++ = XCL_RANGE;
2907                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
2908                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
2909                  *class_utf8data++ = XCL_RANGE;
2910                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
2911                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
2912                  *class_utf8data++ = XCL_RANGE;
2913                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
2914                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
2915                  *class_utf8data++ = XCL_RANGE;
2916                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
2917                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
2918                  *class_utf8data++ = XCL_RANGE;
2919                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
2920                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2921                  }
2922    #endif
2923                continue;
2924                }
2925    
2926              if (-c == ESC_v)
2927                {
2928                SETBIT(classbits, 0x0a); /* LF */
2929                SETBIT(classbits, 0x0b); /* VT */
2930                SETBIT(classbits, 0x0c); /* FF */
2931                SETBIT(classbits, 0x0d); /* CR */
2932                SETBIT(classbits, 0x85); /* NEL */
2933    #ifdef SUPPORT_UTF8
2934                if (utf8)
2935                  {
2936                  class_utf8 = TRUE;
2937                  *class_utf8data++ = XCL_RANGE;
2938                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
2939                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2940                  }
2941    #endif
2942                continue;
2943                }
2944    
2945              if (-c == ESC_V)
2946                {
2947                for (c = 0; c < 32; c++)
2948                  {
2949                  int x = 0xff;
2950                  switch (c)
2951                    {
2952                    case 0x0a/8: x ^= 1 << (0x0a%8);
2953                                 x ^= 1 << (0x0b%8);
2954                                 x ^= 1 << (0x0c%8);
2955                                 x ^= 1 << (0x0d%8);
2956                                 break;
2957                    case 0x85/8: x ^= 1 << (0x85%8); break;
2958                    default: break;
2959                    }
2960                  classbits[c] |= x;
2961                  }
2962    
2963    #ifdef SUPPORT_UTF8
2964                if (utf8)
2965                  {
2966                  class_utf8 = TRUE;
2967                  *class_utf8data++ = XCL_RANGE;
2968                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2969                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
2970                  *class_utf8data++ = XCL_RANGE;
2971                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2972                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2973                  }
2974    #endif
2975                continue;
2976                }
2977    
2978              /* We need to deal with \P and \p in both phases. */
2979    
2980    #ifdef SUPPORT_UCP
2981              if (-c == ESC_p || -c == ESC_P)
2982                {
2983                BOOL negated;
2984                int pdata;
2985                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
2986                if (ptype < 0) goto FAILED;
2987                class_utf8 = TRUE;
2988                *class_utf8data++ = ((-c == ESC_p) != negated)?
2989                  XCL_PROP : XCL_NOTPROP;
2990                *class_utf8data++ = ptype;
2991                *class_utf8data++ = pdata;
2992                class_charcount -= 2;   /* Not a < 256 character */
2993                continue;
2994                }
2995    #endif
2996              /* Unrecognized escapes are faulted if PCRE is running in its
2997              strict mode. By default, for compatibility with Perl, they are
2998              treated as literals. */
2999    
3000              if ((options & PCRE_EXTRA) != 0)
3001                {
3002                *errorcodeptr = ERR7;
3003                goto FAILED;
3004                }
3005    
3006              class_charcount -= 2;  /* Undo the default count from above */
3007              c = *ptr;              /* Get the final character and fall through */
3008              }
3009    
3010            /* Fall through if we have a single character (c >= 0). This may be
3011            greater than 256 in UTF-8 mode. */
3012    
3013            }   /* End of backslash handling */
3014    
3015          /* A single character may be followed by '-' to form a range. However,
3016          Perl does not permit ']' to be the end of the range. A '-' character
3017          at the end is treated as a literal. Perl ignores orphaned \E sequences
3018          entirely. The code for handling \Q and \E is messy. */
3019    
3020          CHECK_RANGE:
3021          while (ptr[1] == '\\' && ptr[2] == 'E')
3022            {
3023            inescq = FALSE;
3024            ptr += 2;
3025            }
3026    
3027          oldptr = ptr;
3028    
3029          /* Remember \r or \n */
3030    
3031          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3032    
3033          /* Check for range */
3034    
3035          if (!inescq && ptr[1] == '-')
3036            {
3037            int d;
3038            ptr += 2;
3039            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
3040    
3041            /* If we hit \Q (not followed by \E) at this point, go into escaped
3042            mode. */
3043    
3044            while (*ptr == '\\' && ptr[1] == 'Q')
3045              {
3046              ptr += 2;
3047              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
3048              inescq = TRUE;
3049              break;
3050              }
3051    
3052            if (*ptr == 0 || (!inescq && *ptr == ']'))
3053              {
3054              ptr = oldptr;
3055              goto LONE_SINGLE_CHARACTER;
3056              }
3057    
3058  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3059          if (utf8)          if (utf8)
# Line 2001  for (;; ptr++) Line 3068  for (;; ptr++)
3068          not any of the other escapes. Perl 5.6 treats a hyphen as a literal          not any of the other escapes. Perl 5.6 treats a hyphen as a literal
3069          in such circumstances. */          in such circumstances. */
3070    
3071          if (d == '\\')          if (!inescq && d == '\\')
3072            {            {
3073            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3074            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3075    
3076            /* \b is backslash; \X is literal X; any other special means the '-'            /* \b is backslash; \X is literal X; \R is literal R; any other
3077            was literal */            special means the '-' was literal */
3078    
3079            if (d < 0)            if (d < 0)
3080              {              {
3081              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = '\b';
3082              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = 'X';
3083                else if (d == -ESC_R) d = 'R'; else
3084                {                {
3085                ptr = oldptr - 2;                ptr = oldptr;
3086                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3087                }                }
3088              }              }
3089            }            }
3090    
3091          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3092          the pre-pass. Optimize one-character ranges */          one-character ranges */
3093    
3094            if (d < c)
3095              {
3096              *errorcodeptr = ERR8;
3097              goto FAILED;
3098              }
3099    
3100          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3101    
3102            /* Remember \r or \n */
3103    
3104            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3105    
3106          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
3107          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3108          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2042  for (;; ptr++) Line 3120  for (;; ptr++)
3120  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3121            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3122              {              {
3123              int occ, ocd;              unsigned int occ, ocd;
3124              int cc = c;              unsigned int cc = c;
3125              int origd = d;              unsigned int origd = d;
3126              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3127                {                {
3128                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3129                      ocd <= (unsigned int)d)
3130                    continue;                          /* Skip embedded ranges */
3131    
3132                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3133                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3134                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3135                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3136                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3137                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3138                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3139                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3140                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3141                  d = ocd;                  d = ocd;
3142                  continue;                  continue;
# Line 2102  for (;; ptr++) Line 3184  for (;; ptr++)
3184          ranges that lie entirely within 0-127 when there is UCP support; else          ranges that lie entirely within 0-127 when there is UCP support; else
3185          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3186    
3187          for (; c <= d; c++)          class_charcount += d - c + 1;
3188            class_lastchar = d;
3189    
3190            /* We can save a bit of time by skipping this in the pre-compile. */
3191    
3192            if (lengthptr == NULL) for (; c <= d; c++)
3193            {            {
3194            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3195            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2110  for (;; ptr++) Line 3197  for (;; ptr++)
3197              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3198              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3199              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3200            }            }
3201    
3202          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2135  for (;; ptr++) Line 3220  for (;; ptr++)
3220  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3221          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3222            {            {
3223            int chartype;            unsigned int othercase;
3224            int othercase;            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3225              {              {
3226              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3227              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2163  for (;; ptr++) Line 3246  for (;; ptr++)
3246          }          }
3247        }        }
3248    
3249      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3250      loop. This "while" is the end of the "do" above. */  
3251        while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
3252    
3253        if (c == 0)                          /* Missing terminating ']' */
3254          {
3255          *errorcodeptr = ERR6;
3256          goto FAILED;
3257          }
3258    
3259    
3260    /* This code has been disabled because it would mean that \s counts as
3261    an explicit \r or \n reference, and that's not really what is wanted. Now
3262    we set the flag only if there is a literal "\r" or "\n" in the class. */
3263    
3264    #if 0
3265        /* Remember whether \r or \n are in this class */
3266    
3267        if (negate_class)
3268          {
3269          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3270          }
3271        else
3272          {
3273          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3274          }
3275    #endif
3276    
     while ((c = *(++ptr)) != ']' || inescq);  
3277    
3278      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
3279      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3280      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3281      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3282      single-bytes only. This is an historical hangover. Maybe one day we can  
3283      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3284        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3285        operate on single-bytes only. This is an historical hangover. Maybe one day
3286        we can tidy these opcodes to handle multi-byte characters.
3287    
3288      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
3289      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2183  for (;; ptr++) Line 3293  for (;; ptr++)
3293      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3294    
3295  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3296      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3297            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3298  #else  #else
3299      if (class_charcount == 1)      if (class_charcount == 1)
3300  #endif  #endif
# Line 2229  for (;; ptr++) Line 3337  for (;; ptr++)
3337      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3338    
3339      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3340      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3341      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3342        the class, so any that were explicitly given as well can be ignored. If
3343        (when there are explicit characters > 255 that must be listed) there are no
3344        characters < 256, we can omit the bitmap in the actual compiled code. */
3345    
3346  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3347      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3348        {        {
3349        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3350        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3351        code += LINK_SIZE;        code += LINK_SIZE;
3352        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3353    
3354        /* If the map is required, install it, and move on to the end of        /* If the map is required, move up the extra data to make room for it;
3355        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3356    
3357        if (class_charcount > 0)        if (class_charcount > 0)
3358          {          {
3359          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3360            memmove(code + 32, code, class_utf8data - code);
3361          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3362          code = class_utf8data;          code = class_utf8data + 32;
         }  
   
       /* If the map is not required, slide down the extra data. */  
   
       else  
         {  
         int len = class_utf8data - (code + 33);  
         memmove(code + 1, code + 33, len);  
         code += len + 1;  
3363          }          }
3364          else code = class_utf8data;
3365    
3366        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3367    
# Line 2266  for (;; ptr++) Line 3370  for (;; ptr++)
3370        }        }
3371  #endif  #endif
3372    
3373      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, set the opcode to OP_CLASS or
3374      and copy it into the code vector. If this is the first thing in the branch,      OP_NCLASS, depending on whether the whole class was negated and whether
3375      there can be no first char setting, whatever the repeat count. Any reqbyte      there were negative specials such as \S in the class. Then copy the 32-byte
3376      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3377    
3378        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3379      if (negate_class)      if (negate_class)
3380        {        {
3381        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3382        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3383        }        }
3384      else      else
3385        {        {
       *code++ = OP_CLASS;  
3386        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3387        }        }
3388      code += 32;      code += 32;
3389      break;      break;
3390    
3391    
3392        /* ===================================================================*/
3393      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3394      has been tested above. */      has been tested above. */
3395    
# Line 2351  for (;; ptr++) Line 3457  for (;; ptr++)
3457        }        }
3458      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3459    
     /* If previous was a recursion, we need to wrap it inside brackets so that  
     it can be replicated if necessary. */  
   
     if (*previous == OP_RECURSE)  
       {  
       memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);  
       code += 1 + LINK_SIZE;  
       *previous = OP_BRA;  
       PUT(previous, 1, code - previous);  
       *code = OP_KET;  
       PUT(code, 1, code - previous);  
       code += 1 + LINK_SIZE;  
       }  
   
3460      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3461      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
3462      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
# Line 2398  for (;; ptr++) Line 3490  for (;; ptr++)
3490          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3491          }          }
3492    
3493          /* If the repetition is unlimited, it pays to see if the next thing on
3494          the line is something that cannot possibly match this character. If so,
3495          automatically possessifying this item gains some performance in the case
3496          where the match fails. */
3497    
3498          if (!possessive_quantifier &&
3499              repeat_max < 0 &&
3500              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3501                options, cd))
3502            {
3503            repeat_type = 0;    /* Force greedy */
3504            possessive_quantifier = TRUE;
3505            }
3506    
3507        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3508        }        }
3509    
3510      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3511      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
3512      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3513      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3514        currently used only for single-byte chars. */
3515    
3516      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3517        {        {
3518        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3519        c = previous[1];        c = previous[1];
3520          if (!possessive_quantifier &&
3521              repeat_max < 0 &&
3522              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3523            {
3524            repeat_type = 0;    /* Force greedy */
3525            possessive_quantifier = TRUE;
3526            }
3527        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3528        }        }
3529    
# Line 2423  for (;; ptr++) Line 3537  for (;; ptr++)
3537      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3538        {        {
3539        uschar *oldcode;        uschar *oldcode;
3540        int prop_type;        int prop_type, prop_value;
3541        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3542        c = *previous;        c = *previous;
3543    
3544          if (!possessive_quantifier &&
3545              repeat_max < 0 &&
3546              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3547            {
3548            repeat_type = 0;    /* Force greedy */
3549            possessive_quantifier = TRUE;
3550            }
3551    
3552        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3553        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3554          previous[1] : -1;          {
3555            prop_type = previous[1];
3556            prop_value = previous[2];
3557            }
3558          else prop_type = prop_value = -1;
3559    
3560        oldcode = code;        oldcode = code;
3561        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2442  for (;; ptr++) Line 3568  for (;; ptr++)
3568        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3569        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3570    
3571        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3572    
3573        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3574    
# Line 2463  for (;; ptr++) Line 3589  for (;; ptr++)
3589          }          }
3590    
3591        /* A repeat minimum of 1 is optimized into some special cases. If the        /* A repeat minimum of 1 is optimized into some special cases. If the
3592        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3593        left in place and, if the maximum is greater than 1, we use OP_UPTO with        left in place and, if the maximum is greater than 1, we use OP_UPTO with
3594        one less than the maximum. */        one less than the maximum. */
3595    
# Line 2490  for (;; ptr++) Line 3616  for (;; ptr++)
3616    
3617          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3618          we have to insert the character for the previous code. For a repeated          we have to insert the character for the previous code. For a repeated
3619          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3620          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3621          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3622    
# Line 2506  for (;; ptr++) Line 3632  for (;; ptr++)
3632  #endif  #endif
3633              {              {
3634              *code++ = c;              *code++ = c;
3635              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3636                  {
3637                  *code++ = prop_type;
3638                  *code++ = prop_value;
3639                  }
3640              }              }
3641            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3642            }            }
3643    
3644          /* Else insert an UPTO if the max is greater than the min, again          /* Else insert an UPTO if the max is greater than the min, again
3645          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3646            UPTO is just for 1 instance, we can use QUERY instead. */
3647    
3648          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3649            {            {
# Line 2525  for (;; ptr++) Line 3656  for (;; ptr++)
3656            else            else
3657  #endif  #endif
3658            *code++ = c;            *code++ = c;
3659            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3660                {
3661                *code++ = prop_type;
3662                *code++ = prop_value;
3663                }
3664            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3665            *code++ = OP_UPTO + repeat_type;  
3666            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3667                {
3668                *code++ = OP_QUERY + repeat_type;
3669                }
3670              else
3671                {
3672                *code++ = OP_UPTO + repeat_type;
3673                PUT2INC(code, 0, repeat_max);
3674                }
3675            }            }
3676          }          }
3677    
# Line 2544  for (;; ptr++) Line 3687  for (;; ptr++)
3687  #endif  #endif
3688        *code++ = c;        *code++ = c;
3689    
3690        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3691        defines the required property. */        define the required property. */
3692    
3693  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3694        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3695            {
3696            *code++ = prop_type;
3697            *code++ = prop_value;
3698            }
3699  #endif  #endif
3700        }        }
3701    
# Line 2571  for (;; ptr++) Line 3718  for (;; ptr++)
3718        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3719        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3720    
3721        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3722    
3723        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3724          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2591  for (;; ptr++) Line 3738  for (;; ptr++)
3738      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
3739      cases. */      cases. */
3740    
3741      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3742               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3743        {        {
3744        register int i;        register int i;
3745        int ketoffset = 0;        int ketoffset = 0;
3746        int len = code - previous;        int len = code - previous;
3747        uschar *bralink = NULL;        uschar *bralink = NULL;
3748    
3749          /* Repeating a DEFINE group is pointless */
3750    
3751          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3752            {
3753            *errorcodeptr = ERR55;
3754            goto FAILED;
3755            }
3756    
3757        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
3758        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
3759        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 2633  for (;; ptr++) Line 3788  for (;; ptr++)
3788          /* If the maximum is 1 or unlimited, we just have to stick in the          /* If the maximum is 1 or unlimited, we just have to stick in the
3789          BRAZERO and do no more at this point. However, we do need to adjust          BRAZERO and do no more at this point. However, we do need to adjust
3790          any OP_RECURSE calls inside the group that refer to the group itself or          any OP_RECURSE calls inside the group that refer to the group itself or
3791          any internal group, because the offset is from the start of the whole          any internal or forward referenced group, because the offset is from
3792          regex. Temporarily terminate the pattern while doing this. */          the start of the whole regex. Temporarily terminate the pattern while
3793            doing this. */
3794    
3795          if (repeat_max <= 1)          if (repeat_max <= 1)
3796            {            {
3797            *code = OP_END;            *code = OP_END;
3798            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3799            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3800            code++;            code++;
3801            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 3813  for (;; ptr++)
3813            {            {
3814            int offset;            int offset;
3815            *code = OP_END;            *code = OP_END;
3816            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3817            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3818            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3819            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2677  for (;; ptr++) Line 3833  for (;; ptr++)
3833        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3834        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3835        copies that we need. If we set a first char from the group, and didn't        copies that we need. If we set a first char from the group, and didn't
3836        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3837          forward reference subroutine calls in the group, there will be entries on
3838          the workspace list; replicate these with an appropriate increment. */
3839    
3840        else        else
3841          {          {
3842          if (repeat_min > 1)          if (repeat_min > 1)
3843            {            {
3844            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3845            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
3846              potential integer overflow. */
3847    
3848              if (lengthptr != NULL)
3849              {              {
3850              memcpy(code, previous, len);              int delta = (repeat_min - 1)*length_prevgroup;
3851              code += len;              if ((double)(repeat_min - 1)*(double)length_prevgroup >
3852                                                                (double)INT_MAX ||
3853                    OFLOW_MAX - *lengthptr < delta)
3854                  {
3855                  *errorcodeptr = ERR20;
3856                  goto FAILED;
3857                  }
3858                *lengthptr += delta;
3859                }
3860    
3861              /* This is compiling for real */
3862    
3863              else
3864                {
3865                if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3866                for (i = 1; i < repeat_min; i++)
3867                  {
3868                  uschar *hc;
3869                  uschar *this_hwm = cd->hwm;
3870                  memcpy(code, previous, len);
3871                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3872                    {
3873                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3874                    cd->hwm += LINK_SIZE;
3875                    }
3876                  save_hwm = this_hwm;
3877                  code += len;
3878                  }
3879              }              }
3880            }            }
3881    
3882          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3883          }          }
3884    
# Line 2697  for (;; ptr++) Line 3886  for (;; ptr++)
3886        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3887        remembering the bracket starts on a stack. In the case of a zero minimum,        remembering the bracket starts on a stack. In the case of a zero minimum,
3888        the first one was set up above. In all cases the repeat_max now specifies        the first one was set up above. In all cases the repeat_max now specifies
3889        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
3890          replicate entries on the forward reference list. */
3891    
3892        if (repeat_max >= 0)        if (repeat_max >= 0)
3893          {          {
3894          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
3895            just adjust the length as if we had. For each repetition we must add 1
3896            to the length for BRAZERO and for all but the last repetition we must
3897            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3898            paranoid checks to avoid integer overflow. */
3899    
3900            if (lengthptr != NULL && repeat_max > 0)
3901              {
3902              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3903                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3904              if ((double)repeat_max *
3905                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3906                      > (double)INT_MAX ||
3907                  OFLOW_MAX - *lengthptr < delta)
3908                {
3909                *errorcodeptr = ERR20;
3910                goto FAILED;
3911                }
3912              *lengthptr += delta;
3913              }
3914    
3915            /* This is compiling for real */
3916    
3917            else for (i = repeat_max - 1; i >= 0; i--)
3918            {            {
3919              uschar *hc;
3920              uschar *this_hwm = cd->hwm;
3921    
3922            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
3923    
3924            /* All but the final copy start a new nesting, maintaining the            /* All but the final copy start a new nesting, maintaining the
# Line 2718  for (;; ptr++) Line 3934  for (;; ptr++)
3934              }              }
3935    
3936            memcpy(code, previous, len);            memcpy(code, previous, len);
3937              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3938                {
3939                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
3940                cd->hwm += LINK_SIZE;
3941                }
3942              save_hwm = this_hwm;
3943            code += len;            code += len;
3944            }            }
3945    
# Line 2740  for (;; ptr++) Line 3962  for (;; ptr++)
3962        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. We
3963        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
3964        don't know if there's been an options resetting after the ket. The        don't know if there's been an options resetting after the ket. The
3965        correct offset was computed above. */        correct offset was computed above.
3966    
3967        else code[-ketoffset] = OP_KETRMAX + repeat_type;        Then, when we are doing the actual compile phase, check to see whether
3968          this group is a non-atomic one that could match an empty string. If so,
3969          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
3970          that runtime checking can be done. [This check is also applied to
3971          atomic groups at runtime, but in a different way.] */
3972    
3973          else
3974            {
3975            uschar *ketcode = code - ketoffset;
3976            uschar *bracode = ketcode - GET(ketcode, 1);
3977            *ketcode = OP_KETRMAX + repeat_type;
3978            if (lengthptr == NULL && *bracode != OP_ONCE)
3979              {
3980              uschar *scode = bracode;
3981              do
3982                {
3983                if (could_be_empty_branch(scode, ketcode, utf8))
3984                  {
3985                  *bracode += OP_SBRA - OP_BRA;
3986                  break;
3987                  }
3988                scode += GET(scode, 1);
3989                }
3990              while (*scode == OP_ALT);
3991              }
3992            }
3993        }        }
3994    
3995      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2753  for (;; ptr++) Line 4000  for (;; ptr++)
4000        goto FAILED;        goto FAILED;
4001        }        }
4002    
4003      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
4004      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
4005      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
4006      which might be the first part of a string whose (former) last char we      anything else, we wrap the entire repeated item inside OP_ONCE brackets.
4007      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
4008        but the special opcodes can optimize it a bit. The repeated item starts at
4009        tempcode, not at previous, which might be the first part of a string whose
4010        (former) last char we repeated.
4011    
4012        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4013        an 'upto' may follow. We skip over an 'exact' item, and then test the
4014        length of what remains before proceeding. */
4015    
4016      if (possessive_quantifier)      if (possessive_quantifier)
4017        {        {
4018        int len = code - tempcode;        int len;
4019        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4020        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
4021        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode];
4022        tempcode[0] = OP_ONCE;        len = code - tempcode;
4023        *code++ = OP_KET;        if (len > 0) switch (*tempcode)
4024        PUTINC(code, 0, len);          {
4025        PUT(tempcode, 1, len);          case OP_STAR:  *tempcode = OP_POSSTAR; break;
4026            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
4027            case OP_QUERY: *tempcode = OP_POSQUERY; break;
4028            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4029    
4030            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4031            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4032            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4033            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4034    
4035            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4036            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4037            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4038            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4039    
4040            default:
4041            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4042            code += 1 + LINK_SIZE;
4043            len += 1 + LINK_SIZE;
4044            tempcode[0] = OP_ONCE;
4045            *code++ = OP_KET;
4046            PUTINC(code, 0, len);
4047            PUT(tempcode, 1, len);
4048            break;
4049            }
4050        }        }
4051    
4052      /* In all case we no longer have a previous item. We also set the      /* In all case we no longer have a previous item. We also set the
# Line 2781  for (;; ptr++) Line 4059  for (;; ptr++)
4059      break;      break;
4060    
4061    
4062      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
4063      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4064      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
4065      of any of them means that this is not a referencing group. They were      parenthesis forms.  */
     checked for validity in the first pass over the string, so we don't have to  
     check for syntax errors here.  */  
4066    
4067      case '(':      case '(':
4068      newoptions = options;      newoptions = options;
4069      skipbytes = 0;      skipbytes = 0;
4070        bravalue = OP_CBRA;
4071        save_hwm = cd->hwm;
4072        reset_bracount = FALSE;
4073    
4074        /* First deal with various "verbs" that can be introduced by '*'. */
4075    
4076        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4077          {
4078          int i, namelen;
4079          const char *vn = verbnames;
4080          const uschar *name = ++ptr;
4081          previous = NULL;
4082          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4083          if (*ptr == ':')
4084            {
4085            *errorcodeptr = ERR59;   /* Not supported */
4086            goto FAILED;
4087            }
4088          if (*ptr != ')')
4089            {
4090            *errorcodeptr = ERR60;
4091            goto FAILED;
4092            }
4093          namelen = ptr - name;
4094          for (i = 0; i < verbcount; i++)
4095            {
4096            if (namelen == verbs[i].len &&
4097                strncmp((char *)name, vn, namelen) == 0)
4098              {
4099              *code = verbs[i].op;
4100              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4101              break;
4102              }
4103            vn += verbs[i].len + 1;
4104            }
4105          if (i < verbcount) continue;
4106          *errorcodeptr = ERR60;
4107          goto FAILED;
4108          }
4109    
4110        /* Deal with the extended parentheses; all are introduced by '?', and the
4111        appearance of any of them means that this is not a capturing group. */
4112    
4113      if (*(++ptr) == '?')      else if (*ptr == '?')
4114        {        {
4115        int set, unset;        int i, set, unset, namelen;
4116        int *optset;        int *optset;
4117          const uschar *name;
4118          uschar *slot;
4119    
4120        switch (*(++ptr))        switch (*(++ptr))
4121          {          {
4122          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
4123          ptr++;          ptr++;
4124          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
4125            if (*ptr == 0)
4126              {
4127              *errorcodeptr = ERR18;
4128              goto FAILED;
4129              }
4130          continue;          continue;
4131    
4132          case ':':                 /* Non-extracting bracket */  
4133            /* ------------------------------------------------------------ */
4134            case '|':                 /* Reset capture count for each branch */
4135            reset_bracount = TRUE;
4136            /* Fall through */
4137    
4138            /* ------------------------------------------------------------ */
4139            case ':':                 /* Non-capturing bracket */
4140          bravalue = OP_BRA;          bravalue = OP_BRA;
4141          ptr++;          ptr++;
4142          break;          break;
4143    
4144    
4145            /* ------------------------------------------------------------ */
4146          case '(':          case '(':
4147          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4148    
4149          /* Condition to test for recursion */          /* A condition can be an assertion, a number (referring to a numbered
4150            group), a name (referring to a named group), or 'R', referring to
4151            recursion. R<digits> and R&name are also permitted for recursion tests.
4152    
4153            There are several syntaxes for testing a named group: (?(name)) is used
4154            by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
4155    
4156            There are two unfortunate ambiguities, caused by history. (a) 'R' can
4157            be the recursive thing or the name 'R' (and similarly for 'R' followed
4158            by digits), and (b) a number could be a name that consists of digits.
4159            In both cases, we look for a name first; if not found, we try the other
4160            cases. */
4161    
4162            /* For conditions that are assertions, check the syntax, and then exit
4163            the switch. This will take control down to where bracketed groups,
4164            including assertions, are processed. */
4165    
4166            if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
4167              break;
4168    
4169            /* Most other conditions use OP_CREF (a couple change to OP_RREF
4170            below), and all need to skip 3 bytes at the start of the group. */
4171    
4172          if (ptr[1] == 'R')          code[1+LINK_SIZE] = OP_CREF;
4173            skipbytes = 3;
4174            refsign = -1;
4175    
4176            /* Check for a test for recursion in a named group. */
4177    
4178            if (ptr[1] == 'R' && ptr[2] == '&')
4179            {            {
4180            code[1+LINK_SIZE] = OP_CREF;            terminator = -1;
4181            PUT2(code, 2+LINK_SIZE, CREF_RECURSE);            ptr += 2;
4182            skipbytes = 3;            code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
           ptr += 3;  
4183            }            }
4184    
4185          /* Condition to test for a numbered subpattern match. We know that          /* Check for a test for a named group's having been set, using the Perl
4186          if a digit follows ( then there will just be digits until ) because          syntax (?(<name>) or (?('name') */
         the syntax was checked in the first pass. */  
4187    
4188          else if ((digitab[ptr[1]] && ctype_digit) != 0)          else if (ptr[1] == '<')
4189            {            {
4190            int condref;                 /* Don't amalgamate; some compilers */            terminator = '>';
           condref = *(++ptr) - '0';    /* grumble at autoincrement in declaration */  
           while (*(++ptr) != ')') condref = condref*10 + *ptr - '0';  
           if (condref == 0)  
             {  
             *errorcodeptr = ERR35;  
             goto FAILED;  
             }  
4191            ptr++;            ptr++;
           code[1+LINK_SIZE] = OP_CREF;  
           PUT2(code, 2+LINK_SIZE, condref);  
           skipbytes = 3;  
4192            }            }
4193          /* For conditions that are assertions, we just fall through, having          else if (ptr[1] == '\'')
4194          set bravalue above. */            {
4195          break;            terminator = '\'';
4196              ptr++;
4197          case '=':                 /* Positive lookahead */            }
4198          bravalue = OP_ASSERT;          else
4199          ptr++;            {
4200          break;            terminator = 0;
4201              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4202              }
4203    
4204          case '!':                 /* Negative lookahead */          /* We now expect to read a name; any thing else is an error */
         bravalue = OP_ASSERT_NOT;  
         ptr++;  
         break;  
4205    
4206          case '<':                 /* Lookbehinds */          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
         switch (*(++ptr))  
4207            {            {
4208            case '=':               /* Positive lookbehind */            ptr += 1;  /* To get the right offset */
4209            bravalue = OP_ASSERTBACK;            *errorcodeptr = ERR28;
4210            ptr++;            goto FAILED;
4211            break;            }
4212    
4213            case '!':               /* Negative lookbehind */          /* Read the name, but also get it as a number if it's all digits */
4214            bravalue = OP_ASSERTBACK_NOT;  
4215            recno = 0;
4216            name = ++ptr;
4217            while ((cd->ctypes[*ptr] & ctype_word) != 0)
4218              {
4219              if (recno >= 0)
4220                recno = ((digitab[*ptr] & ctype_digit) != 0)?
4221                  recno * 10 + *ptr - '0' : -1;
4222            ptr++;            ptr++;
4223            break;            }
4224            namelen = ptr - name;
4225    
4226            if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
4227              {
4228              ptr--;      /* Error offset */
4229              *errorcodeptr = ERR26;
4230              goto FAILED;
4231              }
4232    
4233            /* Do no further checking in the pre-compile phase. */
4234    
4235            if (lengthptr != NULL) break;
4236    
4237            /* In the real compile we do the work of looking for the actual
4238            reference. If the string started with "+" or "-" we require the rest to
4239            be digits, in which case recno will be set. */
4240    
4241            if (refsign > 0)
4242              {
4243              if (recno <= 0)
4244                {
4245                *errorcodeptr = ERR58;
4246                goto FAILED;
4247                }
4248              if (refsign == '-')
4249                {
4250                recno = cd->bracount - recno + 1;
4251                if (recno <= 0)
4252                  {
4253                  *errorcodeptr = ERR15;
4254                  goto FAILED;
4255                  }
4256                }
4257              else recno += cd->bracount;
4258              PUT2(code, 2+LINK_SIZE, recno);
4259              break;
4260              }
4261    
4262            /* Otherwise (did not start with "+" or "-"), start by looking for the
4263            name. */
4264    
4265            slot = cd->name_table;
4266            for (i = 0; i < cd->names_found; i++)
4267              {
4268              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
4269              slot += cd->name_entry_size;
4270              }
4271    
4272            /* Found a previous named subpattern */
4273    
4274            if (i < cd->names_found)
4275              {
4276              recno = GET2(slot, 0);
4277              PUT2(code, 2+LINK_SIZE, recno);
4278              }
4279    
4280            /* Search the pattern for a forward reference */
4281    
4282            else if ((i = find_parens(ptr, cd->bracount, name, namelen,
4283                            (options & PCRE_EXTENDED) != 0)) > 0)
4284              {
4285              PUT2(code, 2+LINK_SIZE, i);
4286              }
4287    
4288            /* If terminator == 0 it means that the name followed directly after
4289            the opening parenthesis [e.g. (?(abc)...] and in this case there are
4290            some further alternatives to try. For the cases where terminator != 0
4291            [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
4292            now checked all the possibilities, so give an error. */
4293    
4294            else if (terminator != 0)
4295              {
4296              *errorcodeptr = ERR15;
4297              goto FAILED;
4298              }
4299    
4300            /* Check for (?(R) for recursion. Allow digits after R to specify a
4301            specific group number. */
4302    
4303            else if (*name == 'R')
4304              {
4305              recno = 0;
4306              for (i = 1; i < namelen; i++)
4307                {
4308                if ((digitab[name[i]] & ctype_digit) == 0)
4309                  {
4310                  *errorcodeptr = ERR15;
4311                  goto FAILED;
4312                  }
4313                recno = recno * 10 + name[i] - '0';
4314                }
4315              if (recno == 0) recno = RREF_ANY;
4316              code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
4317              PUT2(code, 2+LINK_SIZE, recno);
4318              }
4319    
4320            /* Similarly, check for the (?(DEFINE) "condition", which is always
4321            false. */
4322    
4323            else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)
4324              {
4325              code[1+LINK_SIZE] = OP_DEF;
4326              skipbytes = 1;
4327              }
4328    
4329            /* Check for the "name" actually being a subpattern number. */
4330    
4331            else if (recno > 0)
4332              {
4333              PUT2(code, 2+LINK_SIZE, recno);
4334              }
4335    
4336            /* Either an unidentified subpattern, or a reference to (?(0) */
4337    
4338            else
4339              {
4340              *errorcodeptr = (recno == 0)? ERR35: ERR15;
4341              goto FAILED;
4342              }
4343            break;
4344    
4345    
4346            /* ------------------------------------------------------------ */
4347            case '=':                 /* Positive lookahead */
4348            bravalue = OP_ASSERT;
4349            ptr++;
4350            break;
4351    
4352    
4353            /* ------------------------------------------------------------ */
4354            case '!':                 /* Negative lookahead */
4355            ptr++;
4356            if (*ptr == ')')          /* Optimize (?!) */
4357              {
4358              *code++ = OP_FAIL;
4359              previous = NULL;
4360              continue;
4361              }
4362            bravalue = OP_ASSERT_NOT;
4363            break;
4364    
4365    
4366            /* ------------------------------------------------------------ */
4367            case '<':                 /* Lookbehind or named define */
4368            switch (ptr[1])
4369              {
4370              case '=':               /* Positive lookbehind */
4371              bravalue = OP_ASSERTBACK;
4372              ptr += 2;
4373              break;
4374    
4375              case '!':               /* Negative lookbehind */
4376              bravalue = OP_ASSERTBACK_NOT;
4377              ptr += 2;
4378              break;
4379    
4380              default:                /* Could be name define, else bad */
4381              if ((cd->ctypes[ptr[1]] & ctype_word) != 0) goto DEFINE_NAME;
4382              ptr++;                  /* Correct offset for error */
4383              *errorcodeptr = ERR24;
4384              goto FAILED;
4385            }            }
4386          break;          break;
4387    
4388    
4389            /* ------------------------------------------------------------ */
4390          case '>':                 /* One-time brackets */          case '>':                 /* One-time brackets */
4391          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4392          ptr++;          ptr++;
4393          break;          break;
4394    
4395    
4396            /* ------------------------------------------------------------ */
4397          case 'C':                 /* Callout - may be followed by digits; */          case 'C':                 /* Callout - may be followed by digits; */
4398          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4399          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4400          *code++ = OP_CALLOUT;     /* Already checked that the terminating */          *code++ = OP_CALLOUT;
4401            {                       /* closing parenthesis is present. */            {
4402            int n = 0;            int n = 0;
4403            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)