/[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 81 by nigel, Sat Feb 24 21:40:59 2007 UTC revision 268 by ph10, Thu Nov 15 10:28:09 2007 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2007 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56    /* When DEBUG is defined, we need the pcre_printint() function, which is also
57    used by pcretest. DEBUG is not defined when building a production library. */
58    
59    #ifdef DEBUG
60    #include "pcre_printint.src"
61    #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
79    
80  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
81  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
82  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
83  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
84  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
85    so this number is very generous.
86    
87    The same workspace is used during the second, actual compile phase for
88    remembering forward references to groups so that they can be filled in at the
89    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
90    is 4 there is plenty of room. */
91    
92  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
93    
94    
95  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 63  are simple data values; negative values Line 97  are simple data values; negative values
97  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
98  is invalid. */  is invalid. */
99    
100  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */
101  static const short int escapes[] = {  static const short int escapes[] = {
102       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */
103       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */
104     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */
105       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */
106  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */
107  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */
108     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */
109       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */
110  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */
111       0,      0, -ESC_z                                            /* x - z */       0,      0, -ESC_z                                            /* x - z */
112  };  };
113    
114  #else         /* This is the "abnormal" table for EBCDIC systems */  #else           /* This is the "abnormal" table for EBCDIC systems */
115  static const short int escapes[] = {  static const short int escapes[] = {
116  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
117  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 121  static const short int escapes[] = {
121  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
122  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
123  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
124  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
125  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
126  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
127  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
128  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
129  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
130  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
131  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
132  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
133  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
134  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
135  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
136  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
137  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
138  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
144  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
145  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. */
146    
147  static const char *const posix_names[] = {  typedef struct verbitem {
148    "alpha", "lower", "upper",    int   len;
149    "alnum", "ascii", "blank", "cntrl", "digit", "graph",    int   op;
150    "print", "punct", "space", "word",  "xdigit" };  } verbitem;
151    
152    static const char verbnames[] =
153      "ACCEPT\0"
154      "COMMIT\0"
155      "F\0"
156      "FAIL\0"
157      "PRUNE\0"
158      "SKIP\0"
159      "THEN";
160    
161    static verbitem verbs[] = {
162      { 6, OP_ACCEPT },
163      { 6, OP_COMMIT },
164      { 1, OP_FAIL },
165      { 4, OP_FAIL },
166      { 5, OP_PRUNE },
167      { 4, OP_SKIP  },
168      { 4, OP_THEN  }
169    };
170    
171    static int verbcount = sizeof(verbs)/sizeof(verbitem);
172    
173    
174    /* Tables of names of POSIX character classes and their lengths. The names are
175    now all in a single string, to reduce the number of relocations when a shared
176    library is dynamically loaded. The list of lengths is terminated by a zero
177    length entry. The first three must be alpha, lower, upper, as this is assumed
178    for handling case independence. */
179    
180    static const char posix_names[] =
181      "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"
182      "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"
183      "word\0"   "xdigit";
184    
185  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
186    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
187    
188  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
189  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
190  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
191    characters are removed, and for [:alpha:] and [:alnum:] the underscore
192    character is removed. The triples in the table consist of the base map offset,
193    second map offset or -1 if no second map, and a non-negative value for map
194    addition or a negative value for map subtraction (if there are two maps). The
195    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
196    remove vertical space characters, 2 => remove underscore. */
197    
198  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
199    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
200    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
201    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
202    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
203    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
204    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
205    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
206    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
207    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
208    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
209    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
210    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
211    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
212    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
213  };  };
214    
215    
216  /* The texts of compile-time error messages. These are "char *" because they  #define STRING(a)  # a
217  are passed to the outside world. */  #define XSTRING(s) STRING(s)
218    
219  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
220    "no error",  are passed to the outside world. Do not ever re-use any error number, because
221    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
222    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
223    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
224    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
225    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
226    simply count through to the one we want - this isn't a performance issue
227    because these strings are used only when there is a compilation error. */
228    
229    static const char error_texts[] =
230      "no error\0"
231      "\\ at end of pattern\0"
232      "\\c at end of pattern\0"
233      "unrecognized character follows \\\0"
234      "numbers out of order in {} quantifier\0"
235    /* 5 */    /* 5 */
236    "number too big in {} quantifier",    "number too big in {} quantifier\0"
237    "missing terminating ] for character class",    "missing terminating ] for character class\0"
238    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
239    "range out of order in character class",    "range out of order in character class\0"
240    "nothing to repeat",    "nothing to repeat\0"
241    /* 10 */    /* 10 */
242    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
243    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
244    "unrecognized character after (?",    "unrecognized character after (?\0"
245    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
246    "missing )",    "missing )\0"
247    /* 15 */    /* 15 */
248    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
249    "erroffset passed as NULL",    "erroffset passed as NULL\0"
250    "unknown option bit(s) set",    "unknown option bit(s) set\0"
251    "missing ) after comment",    "missing ) after comment\0"
252    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
253    /* 20 */    /* 20 */
254    "regular expression too large",    "regular expression is too large\0"
255    "failed to get memory",    "failed to get memory\0"
256    "unmatched parentheses",    "unmatched parentheses\0"
257    "internal error: code overflow",    "internal error: code overflow\0"
258    "unrecognized character after (?<",    "unrecognized character after (?<\0"
259    /* 25 */    /* 25 */
260    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
261    "malformed number after (?(",    "malformed number or name after (?(\0"
262    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
263    "assertion expected after (?(",    "assertion expected after (?(\0"
264    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
265    /* 30 */    /* 30 */
266    "unknown POSIX class name",    "unknown POSIX class name\0"
267    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
268    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
269    "spare error",    "spare error\0"  /** DEAD **/
270    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
271    /* 35 */    /* 35 */
272    "invalid condition (?(0)",    "invalid condition (?(0)\0"
273    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
274    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
275    "number after (?C is > 255",    "number after (?C is > 255\0"
276    "closing ) for (?C expected",    "closing ) for (?C expected\0"
277    /* 40 */    /* 40 */
278    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
279    "unrecognized character after (?P",    "unrecognized character after (?P\0"
280    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
281    "two named groups have the same name",    "two named subpatterns have the same name\0"
282    "invalid UTF-8 string",    "invalid UTF-8 string\0"
283    /* 45 */    /* 45 */
284    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
285    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
286    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
287  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
288      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
289      /* 50 */
290      "repeated subpattern is too long\0"    /** DEAD **/
291      "octal value is greater than \\377 (not in UTF-8 mode)\0"
292      "internal error: overran compiling workspace\0"
293      "internal error: previously-checked referenced subpattern not found\0"
294      "DEFINE group contains more than one branch\0"
295      /* 55 */
296      "repeating a DEFINE group is not allowed\0"
297      "inconsistent NEWLINE options\0"
298      "\\g is not followed by a braced name or an optionally braced non-zero number\0"
299      "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"
300      "(*VERB) with an argument is not supported\0"
301      /* 60 */
302      "(*VERB) not recognized\0"
303      "number is too big\0"
304      "subpattern name expected after (?&";
305    
306    
307  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 220  For convenience, we use the same bit def Line 320  For convenience, we use the same bit def
320    
321  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
322    
323  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */
324  static const unsigned char digitab[] =  static const unsigned char digitab[] =
325    {    {
326    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 356  static const unsigned char digitab[] =
356    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
357    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
358    
359  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else           /* This is the "abnormal" case, for EBCDIC systems */
360  static const unsigned char digitab[] =  static const unsigned char digitab[] =
361    {    {
362    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 370  static const unsigned char digitab[] =
370    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
371    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
372    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
373    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
374    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
375    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
376    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 404  static const unsigned char ebcdic_charta
404    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
405    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
406    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
407    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
408    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
409    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
410    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 431  static const unsigned char ebcdic_charta
431  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
432    
433  static BOOL  static BOOL
434    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
435      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
436    
437    
438    
439    /*************************************************
440    *            Find an error text                  *
441    *************************************************/
442    
443    /* The error texts are now all in one long string, to save on relocations. As
444    some of the text is of unknown length, we can't use a table of offsets.
445    Instead, just count through the strings. This is not a performance issue
446    because it happens only when there has been a compilation error.
447    
448    Argument:   the error number
449    Returns:    pointer to the error string
450    */
451    
452    static const char *
453    find_error_text(int n)
454    {
455    const char *s = error_texts;
456    for (; n > 0; n--) while (*s++ != 0);
457    return s;
458    }
459    
460    
461  /*************************************************  /*************************************************
# Line 342  static BOOL Line 464  static BOOL
464    
465  /* 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
466  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
467  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
468  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
469  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,
470    ptr is pointing at the \. On exit, it is on the final character of the escape
471    sequence.
472    
473  Arguments:  Arguments:
474    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 479  Arguments:
479    
480  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
481                   negative => a special escape sequence                   negative => a special escape sequence
482                   on error, errorptr is set                   on error, errorcodeptr is set
483  */  */
484    
485  static int  static int
486  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
487    int options, BOOL isclass)    int options, BOOL isclass)
488  {  {
489  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
490    const uschar *ptr = *ptrptr + 1;
491  int c, i;  int c, i;
492    
493    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
494    ptr--;                            /* Set pointer back to the last byte */
495    
496  /* 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. */
497    
 c = *(++ptr);  
498  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
499    
500  /* 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
501  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.
502  Otherwise further processing may be required. */  Otherwise further processing may be required. */
503    
504  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
505  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphameric */
506  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
507    
508  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
509  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */
510  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
511  #endif  #endif
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 515  else if ((i = escapes[c - 0x48]) != 0)
515  else  else
516    {    {
517    const uschar *oldptr;    const uschar *oldptr;
518      BOOL braced, negated;
519    
520    switch (c)    switch (c)
521      {      {
522      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
# Line 401  else Line 530  else
530      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
531      break;      break;
532    
533        /* \g must be followed by a number, either plain or braced. If positive, it
534        is an absolute backreference. If negative, it is a relative backreference.
535        This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a
536        reference to a named group. This is part of Perl's movement towards a
537        unified syntax for back references. As this is synonymous with \k{name}, we
538        fudge it up by pretending it really was \k. */
539    
540        case 'g':
541        if (ptr[1] == '{')
542          {
543          const uschar *p;
544          for (p = ptr+2; *p != 0 && *p != '}'; p++)
545            if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;
546          if (*p != 0 && *p != '}')
547            {
548            c = -ESC_k;
549            break;
550            }
551          braced = TRUE;
552          ptr++;
553          }
554        else braced = FALSE;
555    
556        if (ptr[1] == '-')
557          {
558          negated = TRUE;
559          ptr++;
560          }
561        else negated = FALSE;
562    
563        c = 0;
564        while ((digitab[ptr[1]] & ctype_digit) != 0)
565          c = c * 10 + *(++ptr) - '0';
566    
567        if (c < 0)
568          {
569          *errorcodeptr = ERR61;
570          break;
571          }
572    
573        if (c == 0 || (braced && *(++ptr) != '}'))
574          {
575          *errorcodeptr = ERR57;
576          break;
577          }
578    
579        if (negated)
580          {
581          if (c > bracount)
582            {
583            *errorcodeptr = ERR15;
584            break;
585            }
586          c = bracount - (c - 1);
587          }
588    
589        c = -(ESC_REF + c);
590        break;
591    
592      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
593      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
594      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 422  else Line 610  else
610        c -= '0';        c -= '0';
611        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
612          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
613          if (c < 0)
614            {
615            *errorcodeptr = ERR61;
616            break;
617            }
618        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
619          {          {
620          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 442  else Line 635  else
635        }        }
636    
637      /* \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
638      larger first octal digit. */      larger first octal digit. The original code used just to take the least
639        significant 8 bits of octal numbers (I think this is what early Perls used
640        to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
641        than 3 octal digits. */
642    
643      case '0':      case '0':
644      c -= '0';      c -= '0';
645      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')
646          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - '0';
647      c &= 255;     /* Take least significant 8 bits */      if (!utf8 && c > 255) *errorcodeptr = ERR51;
648      break;      break;
649    
650      /* \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
651      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
652        treated as a data character. */
653    
654      case 'x':      case 'x':
655  #ifdef SUPPORT_UTF8      if (ptr[1] == '{')
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
656        {        {
657        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
658        register int count = 0;        int count = 0;
659    
660        c = 0;        c = 0;
661        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
662          {          {
663          int cc = *pt++;          register int cc = *pt++;
664            if (c == 0 && cc == '0') continue;     /* Leading zeroes */
665          count++;          count++;
666  #if !EBCDIC    /* ASCII coding */  
667    #ifndef EBCDIC  /* ASCII coding */
668          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= 'a') cc -= 32;               /* Convert to upper case */
669          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));
670  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
671          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */
672          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));
673  #endif  #endif
674          }          }
675    
676        if (*pt == '}')        if (*pt == '}')
677          {          {
678          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
679          ptr = pt;          ptr = pt;
680          break;          break;
681          }          }
682    
683        /* 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
684        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
685        }        }
 #endif  
686    
687      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
688    
689      c = 0;      c = 0;
690      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
691        {        {
692        int cc;                               /* Some compilers don't like ++ */        int cc;                               /* Some compilers don't like ++ */
693        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                        /* in initializers */
694  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
695        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= 'a') cc -= 32;              /* Convert to upper case */
696        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));
697  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
698        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= 'z') cc += 64;              /* Convert to upper case */
699        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));
700  #endif  #endif
701        }        }
702      break;      break;
703    
704      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
705        This coding is ASCII-specific, but then the whole concept of \cx is
706        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
707    
708      case 'c':      case 'c':
709      c = *(++ptr);      c = *(++ptr);
710      if (c == 0)      if (c == 0)
711        {        {
712        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
713        return 0;        break;
714        }        }
715    
716      /* 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 */  
717      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= 'a' && c <= 'z') c -= 32;
718      c ^= 0x40;      c ^= 0x40;
719  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
720      if (c >= 'a' && c <= 'z') c += 64;      if (c >= 'a' && c <= 'z') c += 64;
721      c ^= 0xC0;      c ^= 0xC0;
722  #endif  #endif
# Line 560  escape sequence. Line 758  escape sequence.
758  Argument:  Argument:
759    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
760    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
761      dptr           points to an int that is set to the detailed property value
762    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
763    
764  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
765  */  */
766    
767  static int  static int
768  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
769  {  {
770  int c, i, bot, top;  int c, i, bot, top;
771  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
772  char name[4];  char name[32];
773    
774  c = *(++ptr);  c = *(++ptr);
775  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
776    
777  *negptr = FALSE;  *negptr = FALSE;
778    
779  /* \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
780  preceded by ^ for negation. */  negation. */
781    
782  if (c == '{')  if (c == '{')
783    {    {
# Line 587  if (c == '{') Line 786  if (c == '{')
786      *negptr = TRUE;      *negptr = TRUE;
787      ptr++;      ptr++;
788      }      }
789    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
790      {      {
791      c = *(++ptr);      c = *(++ptr);
792      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
793      if (c == '}') break;      if (c == '}') break;
794      name[i] = c;      name[i] = c;
795      }      }
796    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;  
     }  
797    name[i] = 0;    name[i] = 0;
798    }    }
799    
# Line 619  top = _pcre_utt_size; Line 814  top = _pcre_utt_size;
814    
815  while (bot < top)  while (bot < top)
816    {    {
817    i = (bot + top)/2;    i = (bot + top) >> 1;
818    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
819    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
820        {
821        *dptr = _pcre_utt[i].value;
822        return _pcre_utt[i].type;
823        }
824    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
825    }    }
826    
 UNKNOWN_RETURN:  
827  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
828  *ptrptr = ptr;  *ptrptr = ptr;
829  return -1;  return -1;
# Line 741  return p; Line 939  return p;
939    
940    
941  /*************************************************  /*************************************************
942    *       Find forward referenced subpattern       *
943    *************************************************/
944    
945    /* This function scans along a pattern's text looking for capturing
946    subpatterns, and counting them. If it finds a named pattern that matches the
947    name it is given, it returns its number. Alternatively, if the name is NULL, it
948    returns when it reaches a given numbered subpattern. This is used for forward
949    references to subpatterns. We know that if (?P< is encountered, the name will
950    be terminated by '>' because that is checked in the first pass.
951    
952    Arguments:
953      ptr          current position in the pattern
954      count        current count of capturing parens so far encountered
955      name         name to seek, or NULL if seeking a numbered subpattern
956      lorn         name length, or subpattern number if name is NULL
957      xmode        TRUE if we are in /x mode
958    
959    Returns:       the number of the named subpattern, or -1 if not found
960    */
961    
962    static int
963    find_parens(const uschar *ptr, int count, const uschar *name, int lorn,
964      BOOL xmode)
965    {
966    const uschar *thisname;
967    
968    for (; *ptr != 0; ptr++)
969      {
970      int term;
971    
972      /* Skip over backslashed characters and also entire \Q...\E */
973    
974      if (*ptr == '\\')
975        {
976        if (*(++ptr) == 0) return -1;
977        if (*ptr == 'Q') for (;;)
978          {
979          while (*(++ptr) != 0 && *ptr != '\\');
980          if (*ptr == 0) return -1;
981          if (*(++ptr) == 'E') break;
982          }
983        continue;
984        }
985    
986      /* Skip over character classes */
987    
988      if (*ptr == '[')
989        {
990        while (*(++ptr) != ']')
991          {
992          if (*ptr == 0) return -1;
993          if (*ptr == '\\')
994            {
995            if (*(++ptr) == 0) return -1;
996            if (*ptr == 'Q') for (;;)
997              {
998              while (*(++ptr) != 0 && *ptr != '\\');
999              if (*ptr == 0) return -1;
1000              if (*(++ptr) == 'E') break;
1001              }
1002            continue;
1003            }
1004          }
1005        continue;
1006        }
1007    
1008      /* Skip comments in /x mode */
1009    
1010      if (xmode && *ptr == '#')
1011        {
1012        while (*(++ptr) != 0 && *ptr != '\n');
1013        if (*ptr == 0) return -1;
1014        continue;
1015        }
1016    
1017      /* An opening parens must now be a real metacharacter */
1018    
1019      if (*ptr != '(') continue;
1020      if (ptr[1] != '?' && ptr[1] != '*')
1021        {
1022        count++;
1023        if (name == NULL && count == lorn) return count;
1024        continue;
1025        }
1026    
1027      ptr += 2;
1028      if (*ptr == 'P') ptr++;                      /* Allow optional P */
1029    
1030      /* We have to disambiguate (?<! and (?<= from (?<name> */
1031    
1032      if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&
1033           *ptr != '\'')
1034        continue;
1035    
1036      count++;
1037    
1038      if (name == NULL && count == lorn) return count;
1039      term = *ptr++;
1040      if (term == '<') term = '>';
1041      thisname = ptr;
1042      while (*ptr != term) ptr++;
1043      if (name != NULL && lorn == ptr - thisname &&
1044          strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1045        return count;
1046      }
1047    
1048    return -1;
1049    }
1050    
1051    
1052    
1053    /*************************************************
1054  *      Find first significant op code            *  *      Find first significant op code            *
1055  *************************************************/  *************************************************/
1056    
# Line 789  for (;;) Line 1099  for (;;)
1099    
1100      case OP_CALLOUT:      case OP_CALLOUT:
1101      case OP_CREF:      case OP_CREF:
1102      case OP_BRANUMBER:      case OP_RREF:
1103        case OP_DEF:
1104      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1105      break;      break;
1106    
# Line 834  for (;;) Line 1145  for (;;)
1145    {    {
1146    int d;    int d;
1147    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1148    switch (op)    switch (op)
1149      {      {
1150        case OP_CBRA:
1151      case OP_BRA:      case OP_BRA:
1152      case OP_ONCE:      case OP_ONCE:
1153      case OP_COND:      case OP_COND:
1154      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1155      if (d < 0) return d;      if (d < 0) return d;
1156      branchlength += d;      branchlength += d;
1157      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 876  for (;;) Line 1186  for (;;)
1186      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1187    
1188      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1189      case OP_CREF:      case OP_CREF:
1190        case OP_RREF:
1191        case OP_DEF:
1192      case OP_OPT:      case OP_OPT:
1193      case OP_CALLOUT:      case OP_CALLOUT:
1194      case OP_SOD:      case OP_SOD:
# Line 895  for (;;) Line 1206  for (;;)
1206    
1207      case OP_CHAR:      case OP_CHAR:
1208      case OP_CHARNC:      case OP_CHARNC:
1209        case OP_NOT:
1210      branchlength++;      branchlength++;
1211      cc += 2;      cc += 2;
1212  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 921  for (;;) Line 1233  for (;;)
1233    
1234      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1235      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1236        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1237      cc += 4;      cc += 4;
1238      break;      break;
1239    
# Line 928  for (;;) Line 1241  for (;;)
1241    
1242      case OP_PROP:      case OP_PROP:
1243      case OP_NOTPROP:      case OP_NOTPROP:
1244      cc++;      cc += 2;
1245      /* Fall through */      /* Fall through */
1246    
1247      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 1009  Returns:      pointer to the opcode for Line 1322  Returns:      pointer to the opcode for
1322  static const uschar *  static const uschar *
1323  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1324  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1325  for (;;)  for (;;)
1326    {    {
1327    register int c = *code;    register int c = *code;
1328    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1329    else if (c > OP_BRA)  
1330      /* XCLASS is used for classes that cannot be represented just by a bit
1331      map. This includes negated single high-valued characters. The length in
1332      the table is zero; the actual length is stored in the compiled code. */
1333    
1334      if (c == OP_XCLASS) code += GET(code, 1);
1335    
1336      /* Handle capturing bracket */
1337    
1338      else if (c == OP_CBRA)
1339      {      {
1340      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1341      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1342      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1343      }      }
1344    
1345      /* Otherwise, we can get the item's length from the table, except that for
1346      repeated character types, we have to test for \p and \P, which have an extra
1347      two bytes of parameters. */
1348    
1349    else    else
1350      {      {
1351      code += _pcre_OP_lengths[c];      switch(c)
1352          {
1353          case OP_TYPESTAR:
1354          case OP_TYPEMINSTAR:
1355          case OP_TYPEPLUS:
1356          case OP_TYPEMINPLUS:
1357          case OP_TYPEQUERY:
1358          case OP_TYPEMINQUERY:
1359          case OP_TYPEPOSSTAR:
1360          case OP_TYPEPOSPLUS:
1361          case OP_TYPEPOSQUERY:
1362          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1363          break;
1364    
1365  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1366          case OP_TYPEMINUPTO:
1367          case OP_TYPEEXACT:
1368          case OP_TYPEPOSUPTO:
1369          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1370          break;
1371          }
1372    
1373      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* Add in the fixed length from the table */
1374      by a multi-byte character. The length in the table is a minimum, so we have  
1375      to scan along to skip the extra bytes. All opcodes are less than 128, so we      code += _pcre_OP_lengths[c];
1376      can use relatively efficient code. */  
1377      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1378      a multi-byte character. The length in the table is a minimum, so we have to
1379      arrange to skip the extra bytes. */
1380    
1381    #ifdef SUPPORT_UTF8
1382      if (utf8) switch(c)      if (utf8) switch(c)
1383        {        {
1384        case OP_CHAR:        case OP_CHAR:
# Line 1042  for (;;) Line 1386  for (;;)
1386        case OP_EXACT:        case OP_EXACT:
1387        case OP_UPTO:        case OP_UPTO:
1388        case OP_MINUPTO:        case OP_MINUPTO:
1389          case OP_POSUPTO:
1390        case OP_STAR:        case OP_STAR:
1391        case OP_MINSTAR:        case OP_MINSTAR:
1392          case OP_POSSTAR:
1393        case OP_PLUS:        case OP_PLUS:
1394        case OP_MINPLUS:        case OP_MINPLUS:
1395          case OP_POSPLUS:
1396        case OP_QUERY:        case OP_QUERY:
1397        case OP_MINQUERY:        case OP_MINQUERY:
1398        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1399        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;  
1400        break;        break;
1401        }        }
1402  #endif  #endif
# Line 1083  Returns:      pointer to the opcode for Line 1423  Returns:      pointer to the opcode for
1423  static const uschar *  static const uschar *
1424  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1425  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1426  for (;;)  for (;;)
1427    {    {
1428    register int c = *code;    register int c = *code;
1429    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1430    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1431    else if (c > OP_BRA)  
1432      {    /* XCLASS is used for classes that cannot be represented just by a bit
1433      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1434      }    the table is zero; the actual length is stored in the compiled code. */
1435    
1436      if (c == OP_XCLASS) code += GET(code, 1);
1437    
1438      /* Otherwise, we can get the item's length from the table, except that for
1439      repeated character types, we have to test for \p and \P, which have an extra
1440      two bytes of parameters. */
1441    
1442    else    else
1443      {      {
1444      code += _pcre_OP_lengths[c];      switch(c)
1445          {
1446          case OP_TYPESTAR:
1447          case OP_TYPEMINSTAR:
1448          case OP_TYPEPLUS:
1449          case OP_TYPEMINPLUS:
1450          case OP_TYPEQUERY:
1451          case OP_TYPEMINQUERY:
1452          case OP_TYPEPOSSTAR:
1453          case OP_TYPEPOSPLUS:
1454          case OP_TYPEPOSQUERY:
1455          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1456          break;
1457    
1458  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1459          case OP_TYPEUPTO:
1460          case OP_TYPEMINUPTO:
1461          case OP_TYPEEXACT:
1462          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1463          break;
1464          }
1465    
1466        /* Add in the fixed length from the table */
1467    
1468        code += _pcre_OP_lengths[c];
1469    
1470      /* 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
1471      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
1472      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. */  
1473    
1474    #ifdef SUPPORT_UTF8
1475      if (utf8) switch(c)      if (utf8) switch(c)
1476        {        {
1477        case OP_CHAR:        case OP_CHAR:
# Line 1114  for (;;) Line 1479  for (;;)
1479        case OP_EXACT:        case OP_EXACT:
1480        case OP_UPTO:        case OP_UPTO:
1481        case OP_MINUPTO:        case OP_MINUPTO:
1482          case OP_POSUPTO:
1483        case OP_STAR:        case OP_STAR:
1484        case OP_MINSTAR:        case OP_MINSTAR:
1485          case OP_POSSTAR:
1486        case OP_PLUS:        case OP_PLUS:
1487        case OP_MINPLUS:        case OP_MINPLUS:
1488          case OP_POSPLUS:
1489        case OP_QUERY:        case OP_QUERY:
1490        case OP_MINQUERY:        case OP_MINQUERY:
1491        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1492        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;  
1493        break;        break;
1494        }        }
1495  #endif  #endif
# Line 1143  for (;;) Line 1504  for (;;)
1504  *************************************************/  *************************************************/
1505    
1506  /* 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
1507  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()
1508  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
1509  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
1510  whose current branch will already have been scanned.  assertions. If we hit an unclosed bracket, we return "empty" - this means we've
1511    struck an inner bracket whose current branch will already have been scanned.
1512    
1513  Arguments:  Arguments:
1514    code        points to start of search    code        points to start of search
# Line 1160  static BOOL Line 1522  static BOOL
1522  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1523  {  {
1524  register int c;  register int c;
1525  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);
1526       code < endcode;       code < endcode;
1527       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1528    {    {
# Line 1168  for (code = first_significant_code(code Line 1530  for (code = first_significant_code(code
1530    
1531    c = *code;    c = *code;
1532    
1533    if (c >= OP_BRA)    /* Groups with zero repeats can of course be empty; skip them. */
1534    
1535      if (c == OP_BRAZERO || c == OP_BRAMINZERO)
1536        {
1537        code += _pcre_OP_lengths[c];
1538        do code += GET(code, 1); while (*code == OP_ALT);
1539        c = *code;
1540        continue;
1541        }
1542    
1543      /* For other groups, scan the branches. */
1544    
1545      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1546      {      {
1547      BOOL empty_branch;      BOOL empty_branch;
1548      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1184  for (code = first_significant_code(code Line 1558  for (code = first_significant_code(code
1558        }        }
1559      while (*code == OP_ALT);      while (*code == OP_ALT);
1560      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
     code += 1 + LINK_SIZE;  
1561      c = *code;      c = *code;
1562        continue;
1563      }      }
1564    
1565    else switch (c)    /* Handle the other opcodes */
1566    
1567      switch (c)
1568      {      {
1569      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1570        cannot be represented just by a bit map. This includes negated single
1571        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1572        actual length is stored in the compiled code, so we must update "code"
1573        here. */
1574    
1575  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1576      case OP_XCLASS:      case OP_XCLASS:
1577      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1578      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1579  #endif  #endif
1580    
# Line 1244  for (code = first_significant_code(code Line 1624  for (code = first_significant_code(code
1624      case OP_NOT:      case OP_NOT:
1625      case OP_PLUS:      case OP_PLUS:
1626      case OP_MINPLUS:      case OP_MINPLUS:
1627        case OP_POSPLUS:
1628      case OP_EXACT:      case OP_EXACT:
1629      case OP_NOTPLUS:      case OP_NOTPLUS:
1630      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1631        case OP_NOTPOSPLUS:
1632      case OP_NOTEXACT:      case OP_NOTEXACT:
1633      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1634      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1635        case OP_TYPEPOSPLUS:
1636      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1637      return FALSE;      return FALSE;
1638    
1639        /* These are going to continue, as they may be empty, but we have to
1640        fudge the length for the \p and \P cases. */
1641    
1642        case OP_TYPESTAR:
1643        case OP_TYPEMINSTAR:
1644        case OP_TYPEPOSSTAR:
1645        case OP_TYPEQUERY:
1646        case OP_TYPEMINQUERY:
1647        case OP_TYPEPOSQUERY:
1648        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1649        break;
1650    
1651        /* Same for these */
1652    
1653        case OP_TYPEUPTO:
1654        case OP_TYPEMINUPTO:
1655        case OP_TYPEPOSUPTO:
1656        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1657        break;
1658    
1659      /* End of branch */      /* End of branch */
1660    
1661      case OP_KET:      case OP_KET:
# Line 1261  for (code = first_significant_code(code Line 1664  for (code = first_significant_code(code
1664      case OP_ALT:      case OP_ALT:
1665      return TRUE;      return TRUE;
1666    
1667      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1668      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1669    
1670  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1671      case OP_STAR:      case OP_STAR:
1672      case OP_MINSTAR:      case OP_MINSTAR:
1673        case OP_POSSTAR:
1674      case OP_QUERY:      case OP_QUERY:
1675      case OP_MINQUERY:      case OP_MINQUERY:
1676        case OP_POSQUERY:
1677      case OP_UPTO:      case OP_UPTO:
1678      case OP_MINUPTO:      case OP_MINUPTO:
1679        case OP_POSUPTO:
1680      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1681      break;      break;
1682  #endif  #endif
# Line 1366  Returns:     a value representing the na Line 1772  Returns:     a value representing the na
1772  static int  static int
1773  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1774  {  {
1775    const char *pn = posix_names;
1776  register int yield = 0;  register int yield = 0;
1777  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1778    {    {
1779    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1780      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1781      pn += posix_name_lengths[yield] + 1;
1782    yield++;    yield++;
1783    }    }
1784  return -1;  return -1;
# Line 1388  earlier groups that are outside the curr Line 1796  earlier groups that are outside the curr
1796  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
1797  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
1798  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
1799  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,
1800  partially compiled regex must be temporarily terminated with OP_END.  the partially compiled regex must be temporarily terminated with OP_END.
1801    
1802    This function has been extended with the possibility of forward references for
1803    recursions and subroutine calls. It must also check the list of such references
1804    for the group we are dealing with. If it finds that one of the recursions in
1805    the current group is on this list, it adjusts the offset in the list, not the
1806    value in the reference (which is a group number).
1807    
1808  Arguments:  Arguments:
1809    group      points to the start of the group    group      points to the start of the group
1810    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1811    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1812    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1813      save_hwm   the hwm forward reference pointer at the start of the group
1814    
1815  Returns:     nothing  Returns:     nothing
1816  */  */
1817    
1818  static void  static void
1819  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1820      uschar *save_hwm)
1821  {  {
1822  uschar *ptr = group;  uschar *ptr = group;
1823    
1824  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1825    {    {
1826    int offset = GET(ptr, 1);    int offset;
1827    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1828    
1829      /* See if this recursion is on the forward reference list. If so, adjust the
1830      reference. */
1831    
1832      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1833        {
1834        offset = GET(hc, 0);
1835        if (cd->start_code + offset == ptr + 1)
1836          {
1837          PUT(hc, 0, offset + adjust);
1838          break;
1839          }
1840        }
1841    
1842      /* Otherwise, adjust the recursion offset if it's after the start of this
1843      group. */
1844    
1845      if (hc >= cd->hwm)
1846        {
1847        offset = GET(ptr, 1);
1848        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1849        }
1850    
1851    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1852    }    }
1853  }  }
# Line 1486  Yield:        TRUE when range returned; Line 1926  Yield:        TRUE when range returned;
1926  */  */
1927    
1928  static BOOL  static BOOL
1929  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
1930      unsigned int *odptr)
1931  {  {
1932  int c, chartype, othercase, next;  unsigned int c, othercase, next;
1933    
1934  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
1935    {    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
1936    
1937  if (c > d) return FALSE;  if (c > d) return FALSE;
1938    
# Line 1503  next = othercase + 1; Line 1941  next = othercase + 1;
1941    
1942  for (++c; c <= d; c++)  for (++c; c <= d; c++)
1943    {    {
1944    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (_pcre_ucp_othercase(c) != next) break;
         othercase != next)  
     break;  
1945    next++;    next++;
1946    }    }
1947    
# Line 1517  return TRUE; Line 1953  return TRUE;
1953  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
1954    
1955    
1956    
1957  /*************************************************  /*************************************************
1958  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
1959  *************************************************/  *************************************************/
1960    
1961  /* 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
1962  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
1963  bits.  sense to automatically possessify the repeated item.
1964    
1965  Arguments:  Arguments:
1966    optionsptr     pointer to the option bits    op_code       the repeated op code
1967    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
1968    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
1969    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
1970    errorcodeptr   points to error code variable    ptr           next character in pattern
1971    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
1972    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.  
1973    
1974  Returns:         TRUE on success  Returns:        TRUE if possessifying is wanted
                  FALSE, with *errorcodeptr set non-zero on error  
1975  */  */
1976    
1977  static BOOL  static BOOL
1978  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
1979    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const uschar *ptr, int options, compile_data *cd)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
1980  {  {
1981  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];  
1982    
1983  #ifdef SUPPORT_UTF8  /* Skip whitespace and comments in extended mode */
1984  BOOL class_utf8;  
1985  BOOL utf8 = (options & PCRE_UTF8) != 0;  if ((options & PCRE_EXTENDED) != 0)
1986  uschar *class_utf8data;    {
1987      for (;;)
1988        {
1989        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
1990        if (*ptr == '#')
1991          {
1992          while (*(++ptr) != 0)
1993            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
1994          }
1995        else break;
1996        }
1997      }
1998    
1999    /* If the next item is one that we can handle, get its value. A non-negative
2000    value is a character, a negative value is an escape value. */
2001    
2002    if (*ptr == '\\')
2003      {
2004      int temperrorcode = 0;
2005      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2006      if (temperrorcode != 0) return FALSE;
2007      ptr++;    /* Point after the escape sequence */
2008      }
2009    
2010    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2011      {
2012    #ifdef SUPPORT_UTF8
2013      if (utf8) { GETCHARINC(next, ptr); } else
2014    #endif
2015      next = *ptr++;
2016      }
2017    
2018    else return FALSE;
2019    
2020    /* Skip whitespace and comments in extended mode */
2021    
2022    if ((options & PCRE_EXTENDED) != 0)
2023      {
2024      for (;;)
2025        {
2026        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2027        if (*ptr == '#')
2028          {
2029          while (*(++ptr) != 0)
2030            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2031          }
2032        else break;
2033        }
2034      }
2035    
2036    /* If the next thing is itself optional, we have to give up. */
2037    
2038    if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
2039      return FALSE;
2040    
2041    /* Now compare the next item with the previous opcode. If the previous is a
2042    positive single character match, "item" either contains the character or, if
2043    "item" is greater than 127 in utf8 mode, the character's bytes are in
2044    utf8_char. */
2045    
2046    
2047    /* Handle cases when the next item is a character. */
2048    
2049    if (next >= 0) switch(op_code)
2050      {
2051      case OP_CHAR:
2052    #ifdef SUPPORT_UTF8
2053      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2054    #endif
2055      return item != next;
2056    
2057      /* For CHARNC (caseless character) we must check the other case. If we have
2058      Unicode property support, we can use it to test the other case of
2059      high-valued characters. */
2060    
2061      case OP_CHARNC:
2062    #ifdef SUPPORT_UTF8
2063      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2064    #endif
2065      if (item == next) return FALSE;
2066    #ifdef SUPPORT_UTF8
2067      if (utf8)
2068        {
2069        unsigned int othercase;
2070        if (next < 128) othercase = cd->fcc[next]; else
2071    #ifdef SUPPORT_UCP
2072        othercase = _pcre_ucp_othercase((unsigned int)next);
2073    #else
2074        othercase = NOTACHAR;
2075    #endif
2076        return (unsigned int)item != othercase;
2077        }
2078      else
2079    #endif  /* SUPPORT_UTF8 */
2080      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2081    
2082      /* For OP_NOT, "item" must be a single-byte character. */
2083    
2084      case OP_NOT:
2085      if (next < 0) return FALSE;  /* Not a character */
2086      if (item == next) return TRUE;
2087      if ((options & PCRE_CASELESS) == 0) return FALSE;
2088    #ifdef SUPPORT_UTF8
2089      if (utf8)
2090        {
2091        unsigned int othercase;
2092        if (next < 128) othercase = cd->fcc[next]; else
2093    #ifdef SUPPORT_UCP
2094        othercase = _pcre_ucp_othercase(next);
2095    #else
2096        othercase = NOTACHAR;
2097    #endif
2098        return (unsigned int)item == othercase;
2099        }
2100      else
2101    #endif  /* SUPPORT_UTF8 */
2102      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2103    
2104      case OP_DIGIT:
2105      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2106    
2107      case OP_NOT_DIGIT:
2108      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2109    
2110      case OP_WHITESPACE:
2111      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2112    
2113      case OP_NOT_WHITESPACE:
2114      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2115    
2116      case OP_WORDCHAR:
2117      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2118    
2119      case OP_NOT_WORDCHAR:
2120      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2121    
2122      case OP_HSPACE:
2123      case OP_NOT_HSPACE:
2124      switch(next)
2125        {
2126        case 0x09:
2127        case 0x20:
2128        case 0xa0:
2129        case 0x1680:
2130        case 0x180e:
2131        case 0x2000:
2132        case 0x2001:
2133        case 0x2002:
2134        case 0x2003:
2135        case 0x2004:
2136        case 0x2005:
2137        case 0x2006:
2138        case 0x2007:
2139        case 0x2008:
2140        case 0x2009:
2141        case 0x200A:
2142        case 0x202f:
2143        case 0x205f:
2144        case 0x3000:
2145        return op_code != OP_HSPACE;
2146        default:
2147        return op_code == OP_HSPACE;
2148        }
2149    
2150      case OP_VSPACE:
2151      case OP_NOT_VSPACE:
2152      switch(next)
2153        {
2154        case 0x0a:
2155        case 0x0b:
2156        case 0x0c:
2157        case 0x0d:
2158        case 0x85:
2159        case 0x2028:
2160        case 0x2029:
2161        return op_code != OP_VSPACE;
2162        default:
2163        return op_code == OP_VSPACE;
2164        }
2165    
2166      default:
2167      return FALSE;
2168      }
2169    
2170    
2171    /* Handle the case when the next item is \d, \s, etc. */
2172    
2173    switch(op_code)
2174      {
2175      case OP_CHAR:
2176      case OP_CHARNC:
2177    #ifdef SUPPORT_UTF8
2178      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2179    #endif
2180      switch(-next)
2181        {
2182        case ESC_d:
2183        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2184    
2185        case ESC_D:
2186        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2187    
2188        case ESC_s:
2189        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2190    
2191        case ESC_S:
2192        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2193    
2194        case ESC_w:
2195        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2196    
2197        case ESC_W:
2198        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2199    
2200        case ESC_h:
2201        case ESC_H:
2202        switch(item)
2203          {
2204          case 0x09:
2205          case 0x20:
2206          case 0xa0:
2207          case 0x1680:
2208          case 0x180e:
2209          case 0x2000:
2210          case 0x2001:
2211          case 0x2002:
2212          case 0x2003:
2213          case 0x2004:
2214          case 0x2005:
2215          case 0x2006:
2216          case 0x2007:
2217          case 0x2008:
2218          case 0x2009:
2219          case 0x200A:
2220          case 0x202f:
2221          case 0x205f:
2222          case 0x3000:
2223          return -next != ESC_h;
2224          default:
2225          return -next == ESC_h;
2226          }
2227    
2228        case ESC_v:
2229        case ESC_V:
2230        switch(item)
2231          {
2232          case 0x0a:
2233          case 0x0b:
2234          case 0x0c:
2235          case 0x0d:
2236          case 0x85:
2237          case 0x2028:
2238          case 0x2029:
2239          return -next != ESC_v;
2240          default:
2241          return -next == ESC_v;
2242          }
2243    
2244        default:
2245        return FALSE;
2246        }
2247    
2248      case OP_DIGIT:
2249      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2250             next == -ESC_h || next == -ESC_v;
2251    
2252      case OP_NOT_DIGIT:
2253      return next == -ESC_d;
2254    
2255      case OP_WHITESPACE:
2256      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2257    
2258      case OP_NOT_WHITESPACE:
2259      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2260    
2261      case OP_HSPACE:
2262      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2263    
2264      case OP_NOT_HSPACE:
2265      return next == -ESC_h;
2266    
2267      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2268      case OP_VSPACE:
2269      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2270    
2271      case OP_NOT_VSPACE:
2272      return next == -ESC_v;
2273    
2274      case OP_WORDCHAR:
2275      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2276    
2277      case OP_NOT_WORDCHAR:
2278      return next == -ESC_w || next == -ESC_d;
2279    
2280      default:
2281      return FALSE;
2282      }
2283    
2284    /* Control does not reach here */
2285    }
2286    
2287    
2288    
2289    /*************************************************
2290    *           Compile one branch                   *
2291    *************************************************/
2292    
2293    /* Scan the pattern, compiling it into the a vector. If the options are
2294    changed during the branch, the pointer is used to change the external options
2295    bits. This function is used during the pre-compile phase when we are trying
2296    to find out the amount of memory needed, as well as during the real compile
2297    phase. The value of lengthptr distinguishes the two phases.
2298    
2299    Arguments:
2300      optionsptr     pointer to the option bits
2301      codeptr        points to the pointer to the current code point
2302      ptrptr         points to the current pattern pointer
2303      errorcodeptr   points to error code variable
2304      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2305      reqbyteptr     set to the last literal character required, else < 0
2306      bcptr          points to current branch chain
2307      cd             contains pointers to tables etc.
2308      lengthptr      NULL during the real compile phase
2309                     points to length accumulator during pre-compile phase
2310    
2311    Returns:         TRUE on success
2312                     FALSE, with *errorcodeptr set non-zero on error
2313    */
2314    
2315    static BOOL
2316    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2317      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2318      compile_data *cd, int *lengthptr)
2319    {
2320    int repeat_type, op_type;
2321    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2322    int bravalue = 0;
2323    int greedy_default, greedy_non_default;
2324    int firstbyte, reqbyte;
2325    int zeroreqbyte, zerofirstbyte;
2326    int req_caseopt, reqvary, tempreqvary;
2327    int options = *optionsptr;
2328    int after_manual_callout = 0;
2329    int length_prevgroup = 0;
2330    register int c;
2331    register uschar *code = *codeptr;
2332    uschar *last_code = code;
2333    uschar *orig_code = code;
2334    uschar *tempcode;
2335    BOOL inescq = FALSE;
2336    BOOL groupsetfirstbyte = FALSE;
2337    const uschar *ptr = *ptrptr;
2338    const uschar *tempptr;
2339    uschar *previous = NULL;
2340    uschar *previous_callout = NULL;
2341    uschar *save_hwm = NULL;
2342    uschar classbits[32];
2343    
2344    #ifdef SUPPORT_UTF8
2345    BOOL class_utf8;
2346    BOOL utf8 = (options & PCRE_UTF8) != 0;
2347    uschar *class_utf8data;
2348  uschar utf8_char[6];  uschar utf8_char[6];
2349  #else  #else
2350  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2351    uschar *utf8_char = NULL;
2352    #endif
2353    
2354    #ifdef DEBUG
2355    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2356  #endif  #endif
2357    
2358  /* Set up the default and non-default settings for greediness */  /* Set up the default and non-default settings for greediness */
# Line 1604  req_caseopt = ((options & PCRE_CASELESS) Line 2384  req_caseopt = ((options & PCRE_CASELESS)
2384  for (;; ptr++)  for (;; ptr++)
2385    {    {
2386    BOOL negate_class;    BOOL negate_class;
2387      BOOL should_flip_negation;
2388    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2389    BOOL is_quantifier;    BOOL is_quantifier;
2390      BOOL is_recurse;
2391      BOOL reset_bracount;
2392    int class_charcount;    int class_charcount;
2393    int class_lastchar;    int class_lastchar;
2394    int newoptions;    int newoptions;
2395    int recno;    int recno;
2396      int refsign;
2397    int skipbytes;    int skipbytes;
2398    int subreqbyte;    int subreqbyte;
2399    int subfirstbyte;    int subfirstbyte;
2400      int terminator;
2401    int mclength;    int mclength;
2402    uschar mcbuffer[8];    uschar mcbuffer[8];
2403    
2404    /* Next byte in the pattern */    /* Get next byte in the pattern */
2405    
2406    c = *ptr;    c = *ptr;
2407    
2408      /* If we are in the pre-compile phase, accumulate the length used for the
2409      previous cycle of this loop. */
2410    
2411      if (lengthptr != NULL)
2412        {
2413    #ifdef DEBUG
2414        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2415    #endif
2416        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2417          {
2418          *errorcodeptr = ERR52;
2419          goto FAILED;
2420          }
2421    
2422        /* There is at least one situation where code goes backwards: this is the
2423        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2424        the class is simply eliminated. However, it is created first, so we have to
2425        allow memory for it. Therefore, don't ever reduce the length at this point.
2426        */
2427    
2428        if (code < last_code) code = last_code;
2429    
2430        /* Paranoid check for integer overflow */
2431    
2432        if (OFLOW_MAX - *lengthptr < code - last_code)
2433          {
2434          *errorcodeptr = ERR20;
2435          goto FAILED;
2436          }
2437    
2438        *lengthptr += code - last_code;
2439        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2440    
2441        /* If "previous" is set and it is not at the start of the work space, move
2442        it back to there, in order to avoid filling up the work space. Otherwise,
2443        if "previous" is NULL, reset the current code pointer to the start. */
2444    
2445        if (previous != NULL)
2446          {
2447          if (previous > orig_code)
2448            {
2449            memmove(orig_code, previous, code - previous);
2450            code -= previous - orig_code;
2451            previous = orig_code;
2452            }
2453          }
2454        else code = orig_code;
2455    
2456        /* Remember where this code item starts so we can pick up the length
2457        next time round. */
2458    
2459        last_code = code;
2460        }
2461    
2462      /* In the real compile phase, just check the workspace used by the forward
2463      reference list. */
2464    
2465      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2466        {
2467        *errorcodeptr = ERR52;
2468        goto FAILED;
2469        }
2470    
2471    /* 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 */
2472    
2473    if (inescq && c != 0)    if (inescq && c != 0)
# Line 1634  for (;; ptr++) Line 2482  for (;; ptr++)
2482        {        {
2483        if (previous_callout != NULL)        if (previous_callout != NULL)
2484          {          {
2485          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2486              complete_callout(previous_callout, ptr, cd);
2487          previous_callout = NULL;          previous_callout = NULL;
2488          }          }
2489        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1655  for (;; ptr++) Line 2504  for (;; ptr++)
2504    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2505         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2506      {      {
2507      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2508          complete_callout(previous_callout, ptr, cd);
2509      previous_callout = NULL;      previous_callout = NULL;
2510      }      }
2511    
# Line 1666  for (;; ptr++) Line 2516  for (;; ptr++)
2516      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2517      if (c == '#')      if (c == '#')
2518        {        {
2519        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2520        on the Macintosh. */          {
2521        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2522        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2523          if (*ptr != 0) continue;
2524    
2525          /* Else fall through to handle end of string */
2526          c = 0;
2527        }        }
2528      }      }
2529    
# Line 1683  for (;; ptr++) Line 2537  for (;; ptr++)
2537    
2538    switch(c)    switch(c)
2539      {      {
2540      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2541        case 0:                        /* The branch terminates at string end */
2542      case 0:      case '|':                      /* or | or ) */
     case '|':  
2543      case ')':      case ')':
2544      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2545      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2546      *codeptr = code;      *codeptr = code;
2547      *ptrptr = ptr;      *ptrptr = ptr;
2548        if (lengthptr != NULL)
2549          {
2550          if (OFLOW_MAX - *lengthptr < code - last_code)
2551            {
2552            *errorcodeptr = ERR20;
2553            goto FAILED;
2554            }
2555          *lengthptr += code - last_code;   /* To include callout length */
2556          DPRINTF((">> end branch\n"));
2557          }
2558      return TRUE;      return TRUE;
2559    
2560    
2561        /* ===================================================================*/
2562      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2563      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2564    
# Line 1722  for (;; ptr++) Line 2587  for (;; ptr++)
2587      *code++ = OP_ANY;      *code++ = OP_ANY;
2588      break;      break;
2589    
2590      /* Character classes. If the included characters are all < 255 in value, we  
2591      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2592      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
2593      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
2594      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2595        map as usual, then invert it at the end. However, we use a different opcode
2596        so that data characters > 255 can be handled correctly.
2597    
2598      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2599      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 1747  for (;; ptr++) Line 2614  for (;; ptr++)
2614        goto FAILED;        goto FAILED;
2615        }        }
2616    
2617      /* 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,
2618        if the first few characters (either before or after ^) are \Q\E or \E we
2619        skip them too. This makes for compatibility with Perl. */
2620    
2621      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2622        for (;;)
2623        {        {
       negate_class = TRUE;  
2624        c = *(++ptr);        c = *(++ptr);
2625          if (c == '\\')
2626            {
2627            if (ptr[1] == 'E') ptr++;
2628              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2629                else break;
2630            }
2631          else if (!negate_class && c == '^')
2632            negate_class = TRUE;
2633          else break;
2634        }        }
2635      else  
2636        {      /* If a class contains a negative special such as \S, we need to flip the
2637        negate_class = FALSE;      negation flag at the end, so that support for characters > 255 works
2638        }      correctly (they are all included in the class). */
2639    
2640        should_flip_negation = FALSE;
2641    
2642      /* 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
2643      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
2644      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2645    
2646      class_charcount = 0;      class_charcount = 0;
2647      class_lastchar = -1;      class_lastchar = -1;
2648    
2649        /* Initialize the 32-char bit map to all zeros. We build the map in a
2650        temporary bit of memory, in case the class contains only 1 character (less
2651        than 256), because in that case the compiled code doesn't use the bit map.
2652        */
2653    
2654        memset(classbits, 0, 32 * sizeof(uschar));
2655    
2656  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2657      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2658      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2659  #endif  #endif
2660    
     /* 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));  
   
2661      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2662      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
2663      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. */  
2664    
2665      do      if (c != 0) do
2666        {        {
2667          const uschar *oldptr;
2668    
2669  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2670        if (utf8 && c > 127)        if (utf8 && c > 127)
2671          {                           /* Braces are required because the */          {                           /* Braces are required because the */
# Line 1797  for (;; ptr++) Line 2677  for (;; ptr++)
2677    
2678        if (inescq)        if (inescq)
2679          {          {
2680          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2681            {            {
2682            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2683            ptr++;            ptr++;                            /* Skip the 'E' */
2684            continue;            continue;                         /* Carry on with next */
2685            }            }
2686          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2687          }          }
2688    
2689        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1817  for (;; ptr++) Line 2697  for (;; ptr++)
2697            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr, cd))
2698          {          {
2699          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2700          int posix_class, i;          int posix_class, taboffset, tabopt;
2701          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2702            uschar pbits[32];
2703    
2704          if (ptr[1] != ':')          if (ptr[1] != ':')
2705            {            {
# Line 1830  for (;; ptr++) Line 2711  for (;; ptr++)
2711          if (*ptr == '^')          if (*ptr == '^')
2712            {            {
2713            local_negate = TRUE;            local_negate = TRUE;
2714              should_flip_negation = TRUE;  /* Note negative special */
2715            ptr++;            ptr++;
2716            }            }
2717    
# Line 1847  for (;; ptr++) Line 2729  for (;; ptr++)
2729          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
2730            posix_class = 0;            posix_class = 0;
2731    
2732          /* 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
2733          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
2734          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
2735          white space chars afterwards. */          result into the bit map that is being built. */
2736    
2737          posix_class *= 3;          posix_class *= 3;
2738          for (i = 0; i < 3; i++)  
2739            /* Copy in the first table (always present) */
2740    
2741            memcpy(pbits, cbits + posix_class_maps[posix_class],
2742              32 * sizeof(uschar));
2743    
2744            /* If there is a second table, add or remove it as required. */
2745    
2746            taboffset = posix_class_maps[posix_class + 1];
2747            tabopt = posix_class_maps[posix_class + 2];
2748    
2749            if (taboffset >= 0)
2750            {            {
2751            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
2752            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;  
             }  
2753            else            else
2754              {              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;  
             }  
2755            }            }
2756    
2757            /* Not see if we need to remove any special characters. An option
2758            value of 1 removes vertical space and 2 removes underscore. */
2759    
2760            if (tabopt < 0) tabopt = -tabopt;
2761            if (tabopt == 1) pbits[1] &= ~0x3c;
2762              else if (tabopt == 2) pbits[11] &= 0x7f;
2763    
2764            /* Add the POSIX table or its complement into the main table that is
2765            being built and we are done. */
2766    
2767            if (local_negate)
2768              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
2769            else
2770              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
2771    
2772          ptr = tempptr + 1;          ptr = tempptr + 1;
2773          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
2774          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
2775          }          }
2776    
2777        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2778        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
2779        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.
2780        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2781        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  
2782        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2783    
2784        if (c == '\\')        if (c == '\\')
2785          {          {
2786          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2787            if (*errorcodeptr != 0) goto FAILED;
2788    
2789          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */
2790          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 */
2791            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2792          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2793            {            {
2794            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1901  for (;; ptr++) Line 2798  for (;; ptr++)
2798            else inescq = TRUE;            else inescq = TRUE;
2799            continue;            continue;
2800            }            }
2801            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2802    
2803          if (c < 0)          if (c < 0)
2804            {            {
2805            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2806            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2807            switch (-c)  
2808              /* Save time by not doing this in the pre-compile phase. */
2809    
2810              if (lengthptr == NULL) switch (-c)
2811              {              {
2812              case ESC_d:              case ESC_d:
2813              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
2814              continue;              continue;
2815    
2816              case ESC_D:              case ESC_D:
2817                should_flip_negation = TRUE;
2818              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2819              continue;              continue;
2820    
# Line 1921  for (;; ptr++) Line 2823  for (;; ptr++)
2823              continue;              continue;
2824    
2825              case ESC_W:              case ESC_W:
2826                should_flip_negation = TRUE;
2827              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2828              continue;              continue;
2829    
# Line 1930  for (;; ptr++) Line 2833  for (;; ptr++)
2833              continue;              continue;
2834    
2835              case ESC_S:              case ESC_S:
2836                should_flip_negation = TRUE;
2837              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2838              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2839              continue;              continue;
2840    
2841  #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 */  
               }  
2842              continue;              continue;
 #endif  
2843    
2844              /* Unrecognized escapes are faulted if PCRE is running in its              default:    /* Not recognized; fall through */
2845              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 */  
2846              }              }
           }  
   
         /* Fall through if we have a single character (c >= 0). This may be  
         > 256 in UTF-8 mode. */  
2847    
2848          }   /* End of backslash handling */            /* In the pre-compile phase, just do the recognition. */
2849    
2850        /* A single character may be followed by '-' to form a range. However,            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2851        Perl does not permit ']' to be the end of the range. A '-' character                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
       here is treated as a literal. */  
2852    
2853        if (ptr[1] == '-' && ptr[2] != ']')            /* We need to deal with \H, \h, \V, and \v in both phases because
2854          {            they use extra memory. */
         int d;  
         ptr += 2;  
2855    
2856              if (-c == ESC_h)
2857                {
2858                SETBIT(classbits, 0x09); /* VT */
2859                SETBIT(classbits, 0x20); /* SPACE */
2860                SETBIT(classbits, 0xa0); /* NSBP */
2861  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2862          if (utf8)              if (utf8)
2863            {                           /* Braces are required because the */                {
2864            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */                class_utf8 = TRUE;
2865            }                *class_utf8data++ = XCL_SINGLE;
2866                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2867                  *class_utf8data++ = XCL_SINGLE;
2868                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2869                  *class_utf8data++ = XCL_RANGE;
2870                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2871                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2872                  *class_utf8data++ = XCL_SINGLE;
2873                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2874                  *class_utf8data++ = XCL_SINGLE;
2875                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2876                  *class_utf8data++ = XCL_SINGLE;
2877                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2878                  }
2879    #endif
2880                continue;
2881                }
2882    
2883              if (-c == ESC_H)
2884                {
2885                for (c = 0; c < 32; c++)
2886                  {
2887                  int x = 0xff;
2888                  switch (c)
2889                    {
2890                    case 0x09/8: x ^= 1 << (0x09%8); break;
2891                    case 0x20/8: x ^= 1 << (0x20%8); break;
2892                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2893                    default: break;
2894                    }
2895                  classbits[c] |= x;
2896                  }
2897    
2898    #ifdef SUPPORT_UTF8
2899                if (utf8)
2900                  {
2901                  class_utf8 = TRUE;
2902                  *class_utf8data++ = XCL_RANGE;
2903                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2904                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
2905                  *class_utf8data++ = XCL_RANGE;
2906                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
2907                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
2908                  *class_utf8data++ = XCL_RANGE;
2909                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
2910                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
2911                  *class_utf8data++ = XCL_RANGE;
2912                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
2913                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
2914                  *class_utf8data++ = XCL_RANGE;
2915                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
2916                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
2917                  *class_utf8data++ = XCL_RANGE;
2918                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
2919                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
2920                  *class_utf8data++ = XCL_RANGE;
2921                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
2922                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2923                  }
2924    #endif
2925                continue;
2926                }
2927    
2928              if (-c == ESC_v)
2929                {
2930                SETBIT(classbits, 0x0a); /* LF */
2931                SETBIT(classbits, 0x0b); /* VT */
2932                SETBIT(classbits, 0x0c); /* FF */
2933                SETBIT(classbits, 0x0d); /* CR */
2934                SETBIT(classbits, 0x85); /* NEL */
2935    #ifdef SUPPORT_UTF8
2936                if (utf8)
2937                  {
2938                  class_utf8 = TRUE;
2939                  *class_utf8data++ = XCL_RANGE;
2940                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
2941                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2942                  }
2943    #endif
2944                continue;
2945                }
2946    
2947              if (-c == ESC_V)
2948                {
2949                for (c = 0; c < 32; c++)
2950                  {
2951                  int x = 0xff;
2952                  switch (c)
2953                    {
2954                    case 0x0a/8: x ^= 1 << (0x0a%8);
2955                                 x ^= 1 << (0x0b%8);
2956                                 x ^= 1 << (0x0c%8);
2957                                 x ^= 1 << (0x0d%8);
2958                                 break;
2959                    case 0x85/8: x ^= 1 << (0x85%8); break;
2960                    default: break;
2961                    }
2962                  classbits[c] |= x;
2963                  }
2964    
2965    #ifdef SUPPORT_UTF8
2966                if (utf8)
2967                  {
2968                  class_utf8 = TRUE;
2969                  *class_utf8data++ = XCL_RANGE;
2970                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2971                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
2972                  *class_utf8data++ = XCL_RANGE;
2973                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2974                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2975                  }
2976    #endif
2977                continue;
2978                }
2979    
2980              /* We need to deal with \P and \p in both phases. */
2981    
2982    #ifdef SUPPORT_UCP
2983              if (-c == ESC_p || -c == ESC_P)
2984                {
2985                BOOL negated;
2986                int pdata;
2987                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
2988                if (ptype < 0) goto FAILED;
2989                class_utf8 = TRUE;
2990                *class_utf8data++ = ((-c == ESC_p) != negated)?
2991                  XCL_PROP : XCL_NOTPROP;
2992                *class_utf8data++ = ptype;
2993                *class_utf8data++ = pdata;
2994                class_charcount -= 2;   /* Not a < 256 character */
2995                continue;
2996                }
2997    #endif
2998              /* Unrecognized escapes are faulted if PCRE is running in its
2999              strict mode. By default, for compatibility with Perl, they are
3000              treated as literals. */
3001    
3002              if ((options & PCRE_EXTRA) != 0)
3003                {
3004                *errorcodeptr = ERR7;
3005                goto FAILED;
3006                }
3007    
3008              class_charcount -= 2;  /* Undo the default count from above */
3009              c = *ptr;              /* Get the final character and fall through */
3010              }
3011    
3012            /* Fall through if we have a single character (c >= 0). This may be
3013            greater than 256 in UTF-8 mode. */
3014    
3015            }   /* End of backslash handling */
3016    
3017          /* A single character may be followed by '-' to form a range. However,
3018          Perl does not permit ']' to be the end of the range. A '-' character
3019          at the end is treated as a literal. Perl ignores orphaned \E sequences
3020          entirely. The code for handling \Q and \E is messy. */
3021    
3022          CHECK_RANGE:
3023          while (ptr[1] == '\\' && ptr[2] == 'E')
3024            {
3025            inescq = FALSE;
3026            ptr += 2;
3027            }
3028    
3029          oldptr = ptr;
3030    
3031          /* Remember \r or \n */
3032    
3033          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3034    
3035          /* Check for range */
3036    
3037          if (!inescq && ptr[1] == '-')
3038            {
3039            int d;
3040            ptr += 2;
3041            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
3042    
3043            /* If we hit \Q (not followed by \E) at this point, go into escaped
3044            mode. */
3045    
3046            while (*ptr == '\\' && ptr[1] == 'Q')
3047              {
3048              ptr += 2;
3049              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
3050              inescq = TRUE;
3051              break;
3052              }
3053    
3054            if (*ptr == 0 || (!inescq && *ptr == ']'))
3055              {
3056              ptr = oldptr;
3057              goto LONE_SINGLE_CHARACTER;
3058              }
3059    
3060    #ifdef SUPPORT_UTF8
3061            if (utf8)
3062              {                           /* Braces are required because the */
3063              GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */
3064              }
3065          else          else
3066  #endif  #endif
3067          d = *ptr;  /* Not UTF-8 mode */          d = *ptr;  /* Not UTF-8 mode */
# Line 1992  for (;; ptr++) Line 3070  for (;; ptr++)
3070          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
3071          in such circumstances. */          in such circumstances. */
3072    
3073          if (d == '\\')          if (!inescq && d == '\\')
3074            {            {
3075            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3076            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3077    
3078            /* \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
3079            was literal */            special means the '-' was literal */
3080    
3081            if (d < 0)            if (d < 0)
3082              {              {
3083              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = '\b';
3084              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = 'X';
3085                else if (d == -ESC_R) d = 'R'; else
3086                {                {
3087                ptr = oldptr - 2;                ptr = oldptr;
3088                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3089                }                }
3090              }              }
3091            }            }
3092    
3093          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3094          the pre-pass. Optimize one-character ranges */          one-character ranges */
3095    
3096            if (d < c)
3097              {
3098              *errorcodeptr = ERR8;
3099              goto FAILED;
3100              }
3101    
3102          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3103    
3104            /* Remember \r or \n */
3105    
3106            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3107    
3108          /* 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
3109          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3110          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2033  for (;; ptr++) Line 3122  for (;; ptr++)
3122  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3123            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3124              {              {
3125              int occ, ocd;              unsigned int occ, ocd;
3126              int cc = c;              unsigned int cc = c;
3127              int origd = d;              unsigned int origd = d;
3128              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3129                {                {
3130                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3131                      ocd <= (unsigned int)d)
3132                    continue;                          /* Skip embedded ranges */
3133    
3134                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3135                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3136                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3137                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3138                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3139                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3140                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3141                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3142                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3143                  d = ocd;                  d = ocd;
3144                  continue;                  continue;
# Line 2093  for (;; ptr++) Line 3186  for (;; ptr++)
3186          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
3187          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3188    
3189          for (; c <= d; c++)          class_charcount += d - c + 1;
3190            class_lastchar = d;
3191    
3192            /* We can save a bit of time by skipping this in the pre-compile. */
3193    
3194            if (lengthptr == NULL) for (; c <= d; c++)
3195            {            {
3196            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3197            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2101  for (;; ptr++) Line 3199  for (;; ptr++)
3199              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3200              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3201              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3202            }            }
3203    
3204          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2126  for (;; ptr++) Line 3222  for (;; ptr++)
3222  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3223          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3224            {            {
3225            int chartype;            unsigned int othercase;
3226            int othercase;            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3227              {              {
3228              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3229              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2154  for (;; ptr++) Line 3248  for (;; ptr++)
3248          }          }
3249        }        }
3250    
3251      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3252      loop. This "while" is the end of the "do" above. */  
3253        while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
3254    
3255        if (c == 0)                          /* Missing terminating ']' */
3256          {
3257          *errorcodeptr = ERR6;
3258          goto FAILED;
3259          }
3260    
3261    
3262    /* This code has been disabled because it would mean that \s counts as
3263    an explicit \r or \n reference, and that's not really what is wanted. Now
3264    we set the flag only if there is a literal "\r" or "\n" in the class. */
3265    
3266    #if 0
3267        /* Remember whether \r or \n are in this class */
3268    
3269        if (negate_class)
3270          {
3271          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3272          }
3273        else
3274          {
3275          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3276          }
3277    #endif
3278    
     while ((c = *(++ptr)) != ']' || inescq);  
3279    
3280      /* 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
3281      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
3282      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
3283      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3284      single-bytes only. This is an historical hangover. Maybe one day we can  
3285      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3286        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3287        operate on single-bytes only. This is an historical hangover. Maybe one day
3288        we can tidy these opcodes to handle multi-byte characters.
3289    
3290      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
3291      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 2174  for (;; ptr++) Line 3295  for (;; ptr++)
3295      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3296    
3297  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3298      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3299            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3300  #else  #else
3301      if (class_charcount == 1)      if (class_charcount == 1)
3302  #endif  #endif
# Line 2220  for (;; ptr++) Line 3339  for (;; ptr++)
3339      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3340    
3341      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3342      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3343      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3344        the class, so any that were explicitly given as well can be ignored. If
3345        (when there are explicit characters > 255 that must be listed) there are no
3346        characters < 256, we can omit the bitmap in the actual compiled code. */
3347    
3348  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3349      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3350        {        {
3351        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3352        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3353        code += LINK_SIZE;        code += LINK_SIZE;
3354        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3355    
3356        /* 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;
3357        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3358    
3359        if (class_charcount > 0)        if (class_charcount > 0)
3360          {          {
3361          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3362            memmove(code + 32, code, class_utf8data - code);
3363          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3364          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;  
3365          }          }
3366          else code = class_utf8data;
3367    
3368        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3369    
# Line 2257  for (;; ptr++) Line 3372  for (;; ptr++)
3372        }        }
3373  #endif  #endif
3374    
3375      /* 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
3376      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
3377      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
3378      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3379    
3380        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3381      if (negate_class)      if (negate_class)
3382        {        {
3383        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3384        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3385        }        }
3386      else      else
3387        {        {
       *code++ = OP_CLASS;  
3388        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3389        }        }
3390      code += 32;      code += 32;
3391      break;      break;
3392    
3393    
3394        /* ===================================================================*/
3395      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3396      has been tested above. */      has been tested above. */
3397    
# Line 2342  for (;; ptr++) Line 3459  for (;; ptr++)
3459        }        }
3460      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3461    
     /* 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;  
       }  
   
3462      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3463      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
3464      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 2389  for (;; ptr++) Line 3492  for (;; ptr++)
3492          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3493          }          }
3494    
3495          /* If the repetition is unlimited, it pays to see if the next thing on
3496          the line is something that cannot possibly match this character. If so,
3497          automatically possessifying this item gains some performance in the case
3498          where the match fails. */
3499    
3500          if (!possessive_quantifier &&
3501              repeat_max < 0 &&
3502              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3503                options, cd))
3504            {
3505            repeat_type = 0;    /* Force greedy */
3506            possessive_quantifier = TRUE;
3507            }
3508    
3509        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3510        }        }
3511    
3512      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3513      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-
3514      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3515      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3516        currently used only for single-byte chars. */
3517    
3518      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3519        {        {
3520        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3521        c = previous[1];        c = previous[1];
3522          if (!possessive_quantifier &&
3523              repeat_max < 0 &&
3524              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3525            {
3526            repeat_type = 0;    /* Force greedy */
3527            possessive_quantifier = TRUE;
3528            }
3529        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3530        }        }
3531    
# Line 2414  for (;; ptr++) Line 3539  for (;; ptr++)
3539      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3540        {        {
3541        uschar *oldcode;        uschar *oldcode;
3542        int prop_type;        int prop_type, prop_value;
3543        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3544        c = *previous;        c = *previous;
3545    
3546          if (!possessive_quantifier &&
3547              repeat_max < 0 &&
3548              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3549            {
3550            repeat_type = 0;    /* Force greedy */
3551            possessive_quantifier = TRUE;
3552            }
3553    
3554        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3555        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3556          previous[1] : -1;          {
3557            prop_type = previous[1];
3558            prop_value = previous[2];
3559            }
3560          else prop_type = prop_value = -1;
3561    
3562        oldcode = code;        oldcode = code;
3563        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2433  for (;; ptr++) Line 3570  for (;; ptr++)
3570        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3571        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3572    
3573        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3574    
3575        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3576    
# Line 2454  for (;; ptr++) Line 3591  for (;; ptr++)
3591          }          }
3592    
3593        /* 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
3594        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3595        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
3596        one less than the maximum. */        one less than the maximum. */
3597    
# Line 2481  for (;; ptr++) Line 3618  for (;; ptr++)
3618    
3619          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3620          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
3621          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3622          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3623          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3624    
# Line 2497  for (;; ptr++) Line 3634  for (;; ptr++)
3634  #endif  #endif
3635              {              {
3636              *code++ = c;              *code++ = c;
3637              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3638                  {
3639                  *code++ = prop_type;
3640                  *code++ = prop_value;
3641                  }
3642              }              }
3643            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3644            }            }
3645    
3646          /* 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
3647          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3648            UPTO is just for 1 instance, we can use QUERY instead. */
3649    
3650          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3651            {            {
# Line 2516  for (;; ptr++) Line 3658  for (;; ptr++)
3658            else            else
3659  #endif  #endif
3660            *code++ = c;            *code++ = c;
3661            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3662                {
3663                *code++ = prop_type;
3664                *code++ = prop_value;
3665                }
3666            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3667            *code++ = OP_UPTO + repeat_type;  
3668            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3669                {
3670                *code++ = OP_QUERY + repeat_type;
3671                }
3672              else
3673                {
3674                *code++ = OP_UPTO + repeat_type;
3675                PUT2INC(code, 0, repeat_max);
3676                }
3677            }            }
3678          }          }
3679    
# Line 2535  for (;; ptr++) Line 3689  for (;; ptr++)
3689  #endif  #endif
3690        *code++ = c;        *code++ = c;
3691    
3692        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3693        defines the required property. */        define the required property. */
3694    
3695  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3696        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3697            {
3698            *code++ = prop_type;
3699            *code++ = prop_value;
3700            }
3701  #endif  #endif
3702        }        }
3703    
# Line 2562  for (;; ptr++) Line 3720  for (;; ptr++)
3720        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3721        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3722    
3723        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3724    
3725        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3726          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2582  for (;; ptr++) Line 3740  for (;; ptr++)
3740      /* 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
3741      cases. */      cases. */
3742    
3743      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3744               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3745        {        {
3746        register int i;        register int i;
3747        int ketoffset = 0;        int ketoffset = 0;
3748        int len = code - previous;        int len = code - previous;
3749        uschar *bralink = NULL;        uschar *bralink = NULL;
3750    
3751          /* Repeating a DEFINE group is pointless */
3752    
3753          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3754            {
3755            *errorcodeptr = ERR55;
3756            goto FAILED;
3757            }
3758    
3759        /* 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
3760        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
3761        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 2624  for (;; ptr++) Line 3790  for (;; ptr++)
3790          /* 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
3791          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
3792          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
3793          any internal group, because the offset is from the start of the whole          any internal or forward referenced group, because the offset is from
3794          regex. Temporarily terminate the pattern while doing this. */          the start of the whole regex. Temporarily terminate the pattern while
3795            doing this. */
3796    
3797          if (repeat_max <= 1)          if (repeat_max <= 1)
3798            {            {
3799            *code = OP_END;            *code = OP_END;
3800            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3801            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3802            code++;            code++;
3803            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2648  for (;; ptr++) Line 3815  for (;; ptr++)
3815            {            {
3816            int offset;            int offset;
3817            *code = OP_END;            *code = OP_END;
3818            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3819            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3820            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3821            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2668  for (;; ptr++) Line 3835  for (;; ptr++)
3835        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3836        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3837        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
3838        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3839          forward reference subroutine calls in the group, there will be entries on
3840          the workspace list; replicate these with an appropriate increment. */
3841    
3842        else        else
3843          {          {
3844          if (repeat_min > 1)          if (repeat_min > 1)
3845            {            {
3846            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3847            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
3848              potential integer overflow. */
3849    
3850              if (lengthptr != NULL)
3851                {
3852                int delta = (repeat_min - 1)*length_prevgroup;
3853                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3854                                                                (double)INT_MAX ||
3855                    OFLOW_MAX - *lengthptr < delta)
3856                  {
3857                  *errorcodeptr = ERR20;
3858                  goto FAILED;
3859                  }
3860                *lengthptr += delta;
3861                }
3862    
3863              /* This is compiling for real */
3864    
3865              else
3866              {              {
3867              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3868              code += len;              for (i = 1; i < repeat_min; i++)
3869                  {
3870                  uschar *hc;
3871                  uschar *this_hwm = cd->hwm;
3872                  memcpy(code, previous, len);
3873                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3874                    {
3875                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3876                    cd->hwm += LINK_SIZE;
3877                    }
3878                  save_hwm = this_hwm;
3879                  code += len;
3880                  }
3881              }              }
3882            }            }
3883    
3884          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3885          }          }
3886    
# Line 2688  for (;; ptr++) Line 3888  for (;; ptr++)
3888        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3889        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,
3890        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
3891        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
3892          replicate entries on the forward reference list. */
3893    
3894        if (repeat_max >= 0)        if (repeat_max >= 0)
3895          {          {
3896          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
3897            just adjust the length as if we had. For each repetition we must add 1
3898            to the length for BRAZERO and for all but the last repetition we must
3899            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3900            paranoid checks to avoid integer overflow. */
3901    
3902            if (lengthptr != NULL && repeat_max > 0)
3903              {
3904              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3905                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3906              if ((double)repeat_max *
3907                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3908                      > (double)INT_MAX ||
3909                  OFLOW_MAX - *lengthptr < delta)
3910                {
3911                *errorcodeptr = ERR20;
3912                goto FAILED;
3913                }
3914              *lengthptr += delta;
3915              }
3916    
3917            /* This is compiling for real */
3918    
3919            else for (i = repeat_max - 1; i >= 0; i--)
3920            {            {
3921              uschar *hc;
3922              uschar *this_hwm = cd->hwm;
3923    
3924            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
3925    
3926            /* All but the final copy start a new nesting, maintaining the            /* All but the final copy start a new nesting, maintaining the
# Line 2709  for (;; ptr++) Line 3936  for (;; ptr++)
3936              }              }
3937    
3938            memcpy(code, previous, len);            memcpy(code, previous, len);
3939              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3940                {
3941                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
3942                cd->hwm += LINK_SIZE;
3943                }
3944              save_hwm = this_hwm;
3945            code += len;            code += len;
3946            }            }
3947    
# Line 2731  for (;; ptr++) Line 3964  for (;; ptr++)
3964        /* 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
3965        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
3966        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
3967        correct offset was computed above. */        correct offset was computed above.
3968    
3969        else code[-ketoffset] = OP_KETRMAX + repeat_type;        Then, when we are doing the actual compile phase, check to see whether
3970          this group is a non-atomic one that could match an empty string. If so,
3971          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
3972          that runtime checking can be done. [This check is also applied to
3973          atomic groups at runtime, but in a different way.] */
3974    
3975          else
3976            {
3977            uschar *ketcode = code - ketoffset;
3978            uschar *bracode = ketcode - GET(ketcode, 1);
3979            *ketcode = OP_KETRMAX + repeat_type;
3980            if (lengthptr == NULL && *bracode != OP_ONCE)
3981              {
3982              uschar *scode = bracode;
3983              do
3984                {
3985                if (could_be_empty_branch(scode, ketcode, utf8))
3986                  {
3987                  *bracode += OP_SBRA - OP_BRA;
3988                  break;
3989                  }
3990                scode += GET(scode, 1);
3991                }
3992              while (*scode == OP_ALT);
3993              }
3994            }
3995        }        }
3996    
3997      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2744  for (;; ptr++) Line 4002  for (;; ptr++)
4002        goto FAILED;        goto FAILED;
4003        }        }
4004    
4005      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
4006      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
4007      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
4008      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.
4009      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
4010        but the special opcodes can optimize it a bit. The repeated item starts at
4011        tempcode, not at previous, which might be the first part of a string whose
4012        (former) last char we repeated.
4013    
4014        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4015        an 'upto' may follow. We skip over an 'exact' item, and then test the
4016        length of what remains before proceeding. */
4017    
4018      if (possessive_quantifier)      if (possessive_quantifier)
4019        {        {
4020        int len = code - tempcode;        int len;
4021        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4022        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
4023        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode];
4024        tempcode[0] = OP_ONCE;        len = code - tempcode;
4025        *code++ = OP_KET;        if (len > 0) switch (*tempcode)
4026        PUTINC(code, 0, len);          {
4027        PUT(tempcode, 1, len);          case OP_STAR:  *tempcode = OP_POSSTAR; break;
4028            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
4029            case OP_QUERY: *tempcode = OP_POSQUERY; break;
4030            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4031    
4032            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4033            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4034            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4035            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4036    
4037            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4038            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4039            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4040            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4041    
4042            default:
4043            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4044            code += 1 + LINK_SIZE;
4045            len += 1 + LINK_SIZE;
4046            tempcode[0] = OP_ONCE;
4047            *code++ = OP_KET;
4048            PUTINC(code, 0, len);
4049            PUT(tempcode, 1, len);
4050            break;
4051            }
4052        }        }
4053    
4054      /* 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 2772  for (;; ptr++) Line 4061  for (;; ptr++)
4061      break;      break;
4062    
4063    
4064      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
4065      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4066      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
4067      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.  */  
4068    
4069      case '(':      case '(':
4070      newoptions = options;      newoptions = options;
4071      skipbytes = 0;      skipbytes = 0;
4072        bravalue = OP_CBRA;
4073        save_hwm = cd->hwm;
4074        reset_bracount = FALSE;
4075    
4076        /* First deal with various "verbs" that can be introduced by '*'. */
4077    
4078        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4079          {
4080          int i, namelen;
4081          const char *vn = verbnames;
4082          const uschar *name = ++ptr;
4083          previous = NULL;
4084          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4085          if (*ptr == ':')
4086            {
4087            *errorcodeptr = ERR59;   /* Not supported */
4088            goto FAILED;
4089            }
4090          if (*ptr != ')')
4091            {
4092            *errorcodeptr = ERR60;
4093            goto FAILED;
4094            }
4095          namelen = ptr - name;
4096          for (i = 0; i < verbcount; i++)
4097            {
4098            if (namelen == verbs[i].len &&
4099                strncmp((char *)name, vn, namelen) == 0)
4100              {
4101              *code = verbs[i].op;
4102              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4103              break;
4104              }
4105            vn += verbs[i].len + 1;
4106            }
4107          if (i < verbcount) continue;
4108          *errorcodeptr = ERR60;
4109          goto FAILED;
4110          }
4111    
4112        /* Deal with the extended parentheses; all are introduced by '?', and the
4113        appearance of any of them means that this is not a capturing group. */
4114    
4115      if (*(++ptr) == '?')      else if (*ptr == '?')
4116        {        {
4117        int set, unset;        int i, set, unset, namelen;
4118        int *optset;        int *optset;
4119          const uschar *name;
4120          uschar *slot;
4121    
4122        switch (*(++ptr))        switch (*(++ptr))
4123          {          {
4124          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
4125          ptr++;          ptr++;
4126          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
4127            if (*ptr == 0)
4128              {
4129              *errorcodeptr = ERR18;
4130              goto FAILED;
4131              }
4132          continue;          continue;
4133    
4134          case ':':                 /* Non-extracting bracket */  
4135            /* ------------------------------------------------------------ */
4136            case '|':                 /* Reset capture count for each branch */
4137            reset_bracount = TRUE;
4138            /* Fall through */
4139    
4140            /* ------------------------------------------------------------ */
4141            case ':':                 /* Non-capturing bracket */
4142          bravalue = OP_BRA;          bravalue = OP_BRA;
4143          ptr++;          ptr++;
4144          break;          break;
4145    
4146    
4147            /* ------------------------------------------------------------ */
4148          case '(':          case '(':
4149          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4150    
4151          /* Condition to test for recursion */          /* A condition can be an assertion, a number (referring to a numbered
4152            group), a name (referring to a named group), or 'R', referring to
4153            recursion. R<digits> and R&name are also permitted for recursion tests.
4154    
4155            There are several syntaxes for testing a named group: (?(name)) is used
4156            by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
4157    
4158            There are two unfortunate ambiguities, caused by history. (a) 'R' can
4159            be the recursive thing or the name 'R' (and similarly for 'R' followed
4160            by digits), and (b) a number could be a name that consists of digits.
4161            In both cases, we look for a name first; if not found, we try the other
4162            cases. */
4163    
4164            /* For conditions that are assertions, check the syntax, and then exit
4165            the switch. This will take control down to where bracketed groups,
4166            including assertions, are processed. */
4167    
4168            if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
4169              break;
4170    
4171            /* Most other conditions use OP_CREF (a couple change to OP_RREF
4172            below), and all need to skip 3 bytes at the start of the group. */
4173    
4174          if (ptr[1] == 'R')          code[1+LINK_SIZE] = OP_CREF;
4175            skipbytes = 3;
4176            refsign = -1;
4177    
4178            /* Check for a test for recursion in a named group. */
4179    
4180            if (ptr[1] == 'R' && ptr[2] == '&')
4181            {            {
4182            code[1+LINK_SIZE] = OP_CREF;            terminator = -1;
4183            PUT2(code, 2+LINK_SIZE, CREF_RECURSE);            ptr += 2;
4184            skipbytes = 3;            code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
           ptr += 3;  
4185            }            }
4186    
4187          /* 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
4188          if a digit follows ( then there will just be digits until ) because          syntax (?(<name>) or (?('name') */
         the syntax was checked in the first pass. */  
4189    
4190          else if ((digitab[ptr[1]] && ctype_digit) != 0)          else if (ptr[1] == '<')
4191            {            {
4192            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;  
             }  
4193            ptr++;            ptr++;
           code[1+LINK_SIZE] = OP_CREF;  
           PUT2(code, 2+LINK_SIZE, condref);  
           skipbytes = 3;  
4194            }            }
4195          /* For conditions that are assertions, we just fall through, having          else if (ptr[1] == '\'')
4196          set bravalue above. */            {
4197          break;            terminator = '\'';
4198              ptr++;
4199          case '=':                 /* Positive lookahead */            }
4200          bravalue = OP_ASSERT;          else
4201          ptr++;            {
4202          break;            terminator = 0;
4203              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4204              }
4205    
4206          case '!':                 /* Negative lookahead */          /* We now expect to read a name; any thing else is an error */
         bravalue = OP_ASSERT_NOT;  
         ptr++;  
         break;  
4207    
4208          case '<':                 /* Lookbehinds */          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
         switch (*(++ptr))  
4209            {            {
4210            case '=':               /* Positive lookbehind */            ptr += 1;  /* To get the right offset */
4211            bravalue = OP_ASSERTBACK;            *errorcodeptr = ERR28;
4212            ptr++;            goto FAILED;
4213            break;            }
4214    
4215            /* Read the name, but also get it as a number if it's all digits */
4216    
4217            recno = 0;
4218            name = ++ptr;
4219            while ((cd->ctypes[*ptr] & ctype_word) != 0)
4220              {
4221              if (recno >= 0)
4222                recno = ((digitab[*ptr] & ctype_digit) != 0)?
4223                  recno * 10 + *ptr - '0' : -1;
4224              ptr++;
4225              }
4226            namelen = ptr - name;
4227    
4228            if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
4229              {
4230              ptr--;      /* Error offset */
4231              *errorcodeptr = ERR26;
4232              goto FAILED;
4233              }
4234    
4235            /* Do no further checking in the pre-compile phase. */
4236    
4237            if (lengthptr != NULL) break;
4238    
4239            /* In the real compile we do the work of looking for the actual
4240            reference. If the string started with "+" or "-" we require the rest to
4241            be digits, in which case recno will be set. */
4242    
4243            if (refsign > 0)
4244              {
4245              if (recno <= 0)
4246                {
4247                *errorcodeptr = ERR58;
4248                goto FAILED;
4249                }
4250              if (refsign == '-')
4251                {
4252                recno = cd->bracount - recno + 1;
4253                if (recno <= 0)
4254                  {
4255                  *errorcodeptr = ERR15;
4256                  goto FAILED;
4257                  }
4258                }
4259              else recno += cd->bracount;
4260              PUT2(code, 2+LINK_SIZE, recno);
4261              break;
4262              }
4263    
4264            /* Otherwise (did not start with "+" or "-"), start by looking for the
4265            name. */
4266    
4267            slot = cd->name_table;
4268            for (i = 0; i < cd->names_found; i++)
4269              {
4270              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
4271              slot += cd->name_entry_size;
4272              }
4273    
4274            /* Found a previous named subpattern */
4275    
4276            if (i < cd->names_found)
4277              {
4278              recno = GET2(slot, 0);
4279              PUT2(code, 2+LINK_SIZE, recno);
4280              }
4281    
4282            /* Search the pattern for a forward reference */
4283    
4284            else if ((i = find_parens(ptr, cd->bracount, name, namelen,
4285                            (options & PCRE_EXTENDED) != 0)) > 0)
4286              {
4287              PUT2(code, 2+LINK_SIZE, i);
4288              }
4289    
4290            /* If terminator == 0 it means that the name followed directly after
4291            the opening parenthesis [e.g. (?(abc)...] and in this case there are
4292            some further alternatives to try. For the cases where terminator != 0
4293            [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
4294            now checked all the possibilities, so give an error. */
4295    
4296            else if (terminator != 0)
4297              {
4298              *errorcodeptr = ERR15;
4299              goto FAILED;
4300              }
4301    
4302            /* Check for (?(R) for recursion. Allow digits after R to specify a
4303            specific group number. */
4304    
4305            else if (*name == 'R')
4306              {
4307              recno = 0;
4308              for (i = 1; i < namelen; i++)
4309                {
4310                if ((digitab[name[i]] & ctype_digit) == 0)
4311                  {
4312                  *errorcodeptr = ERR15;
4313                  goto FAILED;
4314                  }
4315                recno = recno * 10 + name[i] - '0';
4316                }
4317              if (recno == 0) recno = RREF_ANY;
4318              code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
4319              PUT2(code, 2+LINK_SIZE, recno);
4320              }
4321    
4322            /* Similarly, check for the (?(DEFINE) "condition", which is always
4323            false. */
4324    
4325            else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)
4326              {
4327              code[1+LINK_SIZE] = OP_DEF;
4328              skipbytes = 1;
4329              }
4330    
4331            /* Check for the "name" actually being a subpattern number. */
4332    
4333            else if (recno > 0)
4334              {
4335              PUT2(code, 2+LINK_SIZE, recno);
4336              }
4337    
4338            /* Either an unidentified subpattern, or a reference to (?(0) */
4339    
4340            else
4341              {
4342              *errorcodeptr = (recno == 0)? ERR35: ERR15;
4343              goto FAILED;
4344              }
4345            break;
4346    
4347    
4348            /* ------------------------------------------------------------ */
4349            case '=':                 /* Positive lookahead */
4350            bravalue = OP_ASSERT;
4351            ptr++;
4352            break;
4353    
4354    
4355            /* ------------------------------------------------------------ */
4356            case '!':                 /* Negative lookahead */
4357            ptr++;
4358            if (*ptr == ')')          /* Optimize (?!) */
4359              {
4360              *code++ = OP_FAIL;
4361              previous = NULL;
4362              continue;
4363              }
4364            bravalue = OP_ASSERT_NOT;
4365            break;
4366    
4367    
4368            /* ------------------------------------------------------------ */
4369            case '<':                 /* Lookbehind or named define */
4370            switch (ptr[1])
4371              {
4372              case '=':               /* Positive lookbehind */
4373              bravalue = OP_ASSERTBACK;
4374              ptr += 2;
4375              break;
4376    
4377            case '!':               /* Negative lookbehind */            case '!':               /* Negative lookbehind */
4378            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4379            ptr++;            ptr += 2;
4380            break;            break;
4381    
4382              default:                /* Could be name define, else bad */
4383              if ((cd->ctypes[ptr[1]] & ctype_word) != 0) goto DEFINE_NAME;
4384              ptr++;                  /* Correct offset for error */
4385              *errorcodeptr = ERR24;
4386              goto FAILED;
4387            }            }
4388          break;          break;
4389    
4390    
4391            /* ------------------------------------------------------------ */
4392          case '>':                 /* One-time brackets */          case '>':                 /* One-time brackets */
4393          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4394          ptr++;          ptr++;
4395          break;          break;
4396    
4397    
4398            /* ------------------------------------------------------------ */
4399          case 'C':                 /* Callout - may be followed by digits; */          case 'C':                 /* Callout - may be followed by digits; */
4400          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4401          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4402          *code++ = OP_CALLOUT;     /* Already checked that the terminating */          *code++ = OP_CALLOUT;
4403            {                       /* closing parenthesis is present. */            {
4404            int n = 0;            int n = 0;
4405            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4406              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - '0';
4407              if (*ptr != ')')
4408                {
4409                *errorcodeptr = ERR39;
4410                goto FAILED;
4411                }
4412            if (n > 255)            if (n > 255)
4413              {              {
4414              *errorcodeptr = ERR38;              *errorcodeptr = ERR38;
# Line 2887  for (;; ptr++) Line 4422  for (;; ptr++)
4422          previous = NULL;          previous = NULL;
4423          continue;          continue;
4424    
4425          case 'P':                 /* Named subpattern handling */  
4426          if (*(++ptr) == '<')      /* Definition */          /* ------------------------------------------------------------ */
4427            case 'P':                 /* Python-style named subpattern handling */
4428            if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */
4429              {
4430              is_recurse = *ptr == '>';
4431              terminator = ')';
4432              goto NAMED_REF_OR_RECURSE;
4433              }
4434