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

Diff of /code/trunk/pcre_compile.c

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

revision 79 by nigel, Sat Feb 24 21:40:52 2007 UTC revision 240 by ph10, Tue Sep 11 15:47:20 2007 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2007 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
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";
304    
305    
306  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 220  For convenience, we use the same bit def Line 319  For convenience, we use the same bit def
319    
320  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
321    
322  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */
323  static const unsigned char digitab[] =  static const unsigned char digitab[] =
324    {    {
325    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 355  static const unsigned char digitab[] =
355    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
356    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
357    
358  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else           /* This is the "abnormal" case, for EBCDIC systems */
359  static const unsigned char digitab[] =  static const unsigned char digitab[] =
360    {    {
361    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 369  static const unsigned char digitab[] =
369    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
370    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
371    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
372    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
373    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
374    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
375    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 403  static const unsigned char ebcdic_charta
403    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
404    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
405    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
406    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
407    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
408    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
409    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 430  static const unsigned char ebcdic_charta
430  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
431    
432  static BOOL  static BOOL
433    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
434      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
435    
436    
437    
438    /*************************************************
439    *            Find an error text                  *
440    *************************************************/
441    
442    /* The error texts are now all in one long string, to save on relocations. As
443    some of the text is of unknown length, we can't use a table of offsets.
444    Instead, just count through the strings. This is not a performance issue
445    because it happens only when there has been a compilation error.
446    
447    Argument:   the error number
448    Returns:    pointer to the error string
449    */
450    
451    static const char *
452    find_error_text(int n)
453    {
454    const char *s = error_texts;
455    for (; n > 0; n--) while (*s++ != 0);
456    return s;
457    }
458    
459    
460  /*************************************************  /*************************************************
# Line 342  static BOOL Line 463  static BOOL
463    
464  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
465  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
466  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
467  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
468  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
469    ptr is pointing at the \. On exit, it is on the final character of the escape
470    sequence.
471    
472  Arguments:  Arguments:
473    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 478  Arguments:
478    
479  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
480                   negative => a special escape sequence                   negative => a special escape sequence
481                   on error, errorptr is set                   on error, errorcodeptr is set
482  */  */
483    
484  static int  static int
485  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
486    int options, BOOL isclass)    int options, BOOL isclass)
487  {  {
488  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
489    const uschar *ptr = *ptrptr + 1;
490  int c, i;  int c, i;
491    
492    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
493    ptr--;                            /* Set pointer back to the last byte */
494    
495  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
496    
 c = *(++ptr);  
497  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
498    
499  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in
500  a table. A non-zero result is something that can be returned immediately.  a table. A non-zero result is something that can be returned immediately.
501  Otherwise further processing may be required. */  Otherwise further processing may be required. */
502    
503  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
504  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphameric */
505  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
506    
507  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
508  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */
509  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
510  #endif  #endif
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 514  else if ((i = escapes[c - 0x48]) != 0)
514  else  else
515    {    {
516    const uschar *oldptr;    const uschar *oldptr;
517      BOOL braced, negated;
518    
519    switch (c)    switch (c)
520      {      {
521      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
# Line 401  else Line 529  else
529      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
530      break;      break;
531    
532        /* \g must be followed by a number, either plain or braced. If positive, it
533        is an absolute backreference. If negative, it is a relative backreference.
534        This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a
535        reference to a named group. This is part of Perl's movement towards a
536        unified syntax for back references. As this is synonymous with \k{name}, we
537        fudge it up by pretending it really was \k. */
538    
539        case 'g':
540        if (ptr[1] == '{')
541          {
542          const uschar *p;
543          for (p = ptr+2; *p != 0 && *p != '}'; p++)
544            if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;
545          if (*p != 0 && *p != '}')
546            {
547            c = -ESC_k;
548            break;
549            }
550          braced = TRUE;
551          ptr++;
552          }
553        else braced = FALSE;
554    
555        if (ptr[1] == '-')
556          {
557          negated = TRUE;
558          ptr++;
559          }
560        else negated = FALSE;
561    
562        c = 0;
563        while ((digitab[ptr[1]] & ctype_digit) != 0)
564          c = c * 10 + *(++ptr) - '0';
565    
566        if (c < 0)
567          {
568          *errorcodeptr = ERR61;
569          break;
570          }
571    
572        if (c == 0 || (braced && *(++ptr) != '}'))
573          {
574          *errorcodeptr = ERR57;
575          break;
576          }
577    
578        if (negated)
579          {
580          if (c > bracount)
581            {
582            *errorcodeptr = ERR15;
583            break;
584            }
585          c = bracount - (c - 1);
586          }
587    
588        c = -(ESC_REF + c);
589        break;
590    
591      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
592      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
593      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 422  else Line 609  else
609        c -= '0';        c -= '0';
610        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
611          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
612          if (c < 0)
613            {
614            *errorcodeptr = ERR61;
615            break;
616            }
617        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
618          {          {
619          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 442  else Line 634  else
634        }        }
635    
636      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
637      larger first octal digit. */      larger first octal digit. The original code used just to take the least
638        significant 8 bits of octal numbers (I think this is what early Perls used
639        to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
640        than 3 octal digits. */
641    
642      case '0':      case '0':
643      c -= '0';      c -= '0';
644      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')
645          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - '0';
646      c &= 255;     /* Take least significant 8 bits */      if (!utf8 && c > 255) *errorcodeptr = ERR51;
647      break;      break;
648    
649      /* \x is complicated when UTF-8 is enabled. \x{ddd} is a character number      /* \x is complicated. \x{ddd} is a character number which can be greater
650      which can be greater than 0xff, but only if the ddd are hex digits. */      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
651        treated as a data character. */
652    
653      case 'x':      case 'x':
654  #ifdef SUPPORT_UTF8      if (ptr[1] == '{')
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
655        {        {
656        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
657        register int count = 0;        int count = 0;
658    
659        c = 0;        c = 0;
660        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
661          {          {
662          int cc = *pt++;          register int cc = *pt++;
663            if (c == 0 && cc == '0') continue;     /* Leading zeroes */
664          count++;          count++;
665  #if !EBCDIC    /* ASCII coding */  
666    #ifndef EBCDIC  /* ASCII coding */
667          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= 'a') cc -= 32;               /* Convert to upper case */
668          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));
669  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
670          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */
671          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));
672  #endif  #endif
673          }          }
674    
675        if (*pt == '}')        if (*pt == '}')
676          {          {
677          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
678          ptr = pt;          ptr = pt;
679          break;          break;
680          }          }
681    
682        /* If the sequence of hex digits does not end with '}', then we don't        /* If the sequence of hex digits does not end with '}', then we don't
683        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
684        }        }
 #endif  
685    
686      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
687    
688      c = 0;      c = 0;
689      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
690        {        {
691        int cc;                               /* Some compilers don't like ++ */        int cc;                               /* Some compilers don't like ++ */
692        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                        /* in initializers */
693  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
694        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= 'a') cc -= 32;              /* Convert to upper case */
695        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));
696  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
697        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= 'z') cc += 64;              /* Convert to upper case */
698        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));
699  #endif  #endif
700        }        }
701      break;      break;
702    
703      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
704        This coding is ASCII-specific, but then the whole concept of \cx is
705        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
706    
707      case 'c':      case 'c':
708      c = *(++ptr);      c = *(++ptr);
709      if (c == 0)      if (c == 0)
710        {        {
711        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
712        return 0;        break;
713        }        }
714    
715      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII coding */
     is ASCII-specific, but then the whole concept of \cx is ASCII-specific.  
     (However, an EBCDIC equivalent has now been added.) */  
   
 #if !EBCDIC    /* ASCII coding */  
716      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= 'a' && c <= 'z') c -= 32;
717      c ^= 0x40;      c ^= 0x40;
718  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
719      if (c >= 'a' && c <= 'z') c += 64;      if (c >= 'a' && c <= 'z') c += 64;
720      c ^= 0xC0;      c ^= 0xC0;
721  #endif  #endif
# Line 560  escape sequence. Line 757  escape sequence.
757  Argument:  Argument:
758    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
759    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
760      dptr           points to an int that is set to the detailed property value
761    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
762    
763  Returns:     value from ucp_type_table, or -1 for an invalid type  Returns:         type value from ucp_type_table, or -1 for an invalid type
764  */  */
765    
766  static int  static int
767  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
768  {  {
769  int c, i, bot, top;  int c, i, bot, top;
770  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
771  char name[4];  char name[32];
772    
773  c = *(++ptr);  c = *(++ptr);
774  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
775    
776  *negptr = FALSE;  *negptr = FALSE;
777    
778  /* \P or \p can be followed by a one- or two-character name in {}, optionally  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
779  preceded by ^ for negation. */  negation. */
780    
781  if (c == '{')  if (c == '{')
782    {    {
# Line 587  if (c == '{') Line 785  if (c == '{')
785      *negptr = TRUE;      *negptr = TRUE;
786      ptr++;      ptr++;
787      }      }
788    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
789      {      {
790      c = *(++ptr);      c = *(++ptr);
791      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
792      if (c == '}') break;      if (c == '}') break;
793      name[i] = c;      name[i] = c;
794      }      }
795    if (c !='}')   /* Try to distinguish error cases */    if (c !='}') goto ERROR_RETURN;
     {  
     while (*(++ptr) != 0 && *ptr != '}');  
     if (*ptr == '}') goto UNKNOWN_RETURN; else goto ERROR_RETURN;  
     }  
796    name[i] = 0;    name[i] = 0;
797    }    }
798    
# Line 619  top = _pcre_utt_size; Line 813  top = _pcre_utt_size;
813    
814  while (bot < top)  while (bot < top)
815    {    {
816    i = (bot + top)/2;    i = (bot + top) >> 1;
817    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
818    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
819        {
820        *dptr = _pcre_utt[i].value;
821        return _pcre_utt[i].type;
822        }
823    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
824    }    }
825    
 UNKNOWN_RETURN:  
826  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
827  *ptrptr = ptr;  *ptrptr = ptr;
828  return -1;  return -1;
# Line 698  read_repeat_counts(const uschar *p, int Line 895  read_repeat_counts(const uschar *p, int
895  int min = 0;  int min = 0;
896  int max = -1;  int max = -1;
897    
898    /* Read the minimum value and do a paranoid check: a negative value indicates
899    an integer overflow. */
900    
901  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';
902    if (min < 0 || min > 65535)
903      {
904      *errorcodeptr = ERR5;
905      return p;
906      }
907    
908    /* Read the maximum value if there is one, and again do a paranoid on its size.
909    Also, max must not be less than min. */
910    
911  if (*p == '}') max = min; else  if (*p == '}') max = min; else
912    {    {
# Line 706  if (*p == '}') max = min; else Line 914  if (*p == '}') max = min; else
914      {      {
915      max = 0;      max = 0;
916      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';
917        if (max < 0 || max > 65535)
918          {
919          *errorcodeptr = ERR5;
920          return p;
921          }
922      if (max < min)      if (max < min)
923        {        {
924        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 927  if (*p == '}') max = min; else
927      }      }
928    }    }
929    
930  /* Do paranoid checks, then fill in the required variables, and pass back the  /* Fill in the required variables, and pass back the pointer to the terminating
931  pointer to the terminating '}'. */  '}'. */
932    
933  if (min > 65535 || max > 65535)  *minp = min;
934    *errorcodeptr = ERR5;  *maxp = max;
935  else  return p;
936    }
937    
938    
939    
940    /*************************************************
941    *       Find forward referenced subpattern       *
942    *************************************************/
943    
944    /* This function scans along a pattern's text looking for capturing
945    subpatterns, and counting them. If it finds a named pattern that matches the
946    name it is given, it returns its number. Alternatively, if the name is NULL, it
947    returns when it reaches a given numbered subpattern. This is used for forward
948    references to subpatterns. We know that if (?P< is encountered, the name will
949    be terminated by '>' because that is checked in the first pass.
950    
951    Arguments:
952      ptr          current position in the pattern
953      count        current count of capturing parens so far encountered
954      name         name to seek, or NULL if seeking a numbered subpattern
955      lorn         name length, or subpattern number if name is NULL
956      xmode        TRUE if we are in /x mode
957    
958    Returns:       the number of the named subpattern, or -1 if not found
959    */
960    
961    static int
962    find_parens(const uschar *ptr, int count, const uschar *name, int lorn,
963      BOOL xmode)
964    {
965    const uschar *thisname;
966    
967    for (; *ptr != 0; ptr++)
968    {    {
969    *minp = min;    int term;
970    *maxp = max;  
971      /* Skip over backslashed characters and also entire \Q...\E */
972    
973      if (*ptr == '\\')
974        {
975        if (*(++ptr) == 0) return -1;
976        if (*ptr == 'Q') for (;;)
977          {
978          while (*(++ptr) != 0 && *ptr != '\\');
979          if (*ptr == 0) return -1;
980          if (*(++ptr) == 'E') break;
981          }
982        continue;
983        }
984    
985      /* Skip over character classes */
986    
987      if (*ptr == '[')
988        {
989        while (*(++ptr) != ']')
990          {
991          if (*ptr == 0) return -1;
992          if (*ptr == '\\')
993            {
994            if (*(++ptr) == 0) return -1;
995            if (*ptr == 'Q') for (;;)
996              {
997              while (*(++ptr) != 0 && *ptr != '\\');
998              if (*ptr == 0) return -1;
999              if (*(++ptr) == 'E') break;
1000              }
1001            continue;
1002            }
1003          }
1004        continue;
1005        }
1006    
1007      /* Skip comments in /x mode */
1008    
1009      if (xmode && *ptr == '#')
1010        {
1011        while (*(++ptr) != 0 && *ptr != '\n');
1012        if (*ptr == 0) return -1;
1013        continue;
1014        }
1015    
1016      /* An opening parens must now be a real metacharacter */
1017    
1018      if (*ptr != '(') continue;
1019      if (ptr[1] != '?' && ptr[1] != '*')
1020        {
1021        count++;
1022        if (name == NULL && count == lorn) return count;
1023        continue;
1024        }
1025    
1026      ptr += 2;
1027      if (*ptr == 'P') ptr++;                      /* Allow optional P */
1028    
1029      /* We have to disambiguate (?<! and (?<= from (?<name> */
1030    
1031      if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&
1032           *ptr != '\'')
1033        continue;
1034    
1035      count++;
1036    
1037      if (name == NULL && count == lorn) return count;
1038      term = *ptr++;
1039      if (term == '<') term = '>';
1040      thisname = ptr;
1041      while (*ptr != term) ptr++;
1042      if (name != NULL && lorn == ptr - thisname &&
1043          strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1044        return count;
1045    }    }
1046  return p;  
1047    return -1;
1048  }  }
1049    
1050    
# Line 778  for (;;) Line 1098  for (;;)
1098    
1099      case OP_CALLOUT:      case OP_CALLOUT:
1100      case OP_CREF:      case OP_CREF:
1101      case OP_BRANUMBER:      case OP_RREF:
1102        case OP_DEF:
1103      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1104      break;      break;
1105    
# Line 823  for (;;) Line 1144  for (;;)
1144    {    {
1145    int d;    int d;
1146    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1147    switch (op)    switch (op)
1148      {      {
1149        case OP_CBRA:
1150      case OP_BRA:      case OP_BRA:
1151      case OP_ONCE:      case OP_ONCE:
1152      case OP_COND:      case OP_COND:
1153      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1154      if (d < 0) return d;      if (d < 0) return d;
1155      branchlength += d;      branchlength += d;
1156      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 865  for (;;) Line 1185  for (;;)
1185      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1186    
1187      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1188      case OP_CREF:      case OP_CREF:
1189        case OP_RREF:
1190        case OP_DEF:
1191      case OP_OPT:      case OP_OPT:
1192      case OP_CALLOUT:      case OP_CALLOUT:
1193      case OP_SOD:      case OP_SOD:
# Line 884  for (;;) Line 1205  for (;;)
1205    
1206      case OP_CHAR:      case OP_CHAR:
1207      case OP_CHARNC:      case OP_CHARNC:
1208        case OP_NOT:
1209      branchlength++;      branchlength++;
1210      cc += 2;      cc += 2;
1211  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 910  for (;;) Line 1232  for (;;)
1232    
1233      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1234      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1235        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1236      cc += 4;      cc += 4;
1237      break;      break;
1238    
# Line 917  for (;;) Line 1240  for (;;)
1240    
1241      case OP_PROP:      case OP_PROP:
1242      case OP_NOTPROP:      case OP_NOTPROP:
1243      cc++;      cc += 2;
1244      /* Fall through */      /* Fall through */
1245    
1246      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 998  Returns:      pointer to the opcode for Line 1321  Returns:      pointer to the opcode for
1321  static const uschar *  static const uschar *
1322  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1323  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1324  for (;;)  for (;;)
1325    {    {
1326    register int c = *code;    register int c = *code;
1327    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1328    else if (c > OP_BRA)  
1329      /* XCLASS is used for classes that cannot be represented just by a bit
1330      map. This includes negated single high-valued characters. The length in
1331      the table is zero; the actual length is stored in the compiled code. */
1332    
1333      if (c == OP_XCLASS) code += GET(code, 1);
1334    
1335      /* Handle capturing bracket */
1336    
1337      else if (c == OP_CBRA)
1338      {      {
1339      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1340      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1341      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1342      }      }
1343    
1344      /* Otherwise, we can get the item's length from the table, except that for
1345      repeated character types, we have to test for \p and \P, which have an extra
1346      two bytes of parameters. */
1347    
1348    else    else
1349      {      {
1350      code += _pcre_OP_lengths[c];      switch(c)
1351          {
1352          case OP_TYPESTAR:
1353          case OP_TYPEMINSTAR:
1354          case OP_TYPEPLUS:
1355          case OP_TYPEMINPLUS:
1356          case OP_TYPEQUERY:
1357          case OP_TYPEMINQUERY:
1358          case OP_TYPEPOSSTAR:
1359          case OP_TYPEPOSPLUS:
1360          case OP_TYPEPOSQUERY:
1361          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1362          break;
1363    
1364  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1365          case OP_TYPEMINUPTO:
1366          case OP_TYPEEXACT:
1367          case OP_TYPEPOSUPTO:
1368          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1369          break;
1370          }
1371    
1372      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* Add in the fixed length from the table */
1373      by a multi-byte character. The length in the table is a minimum, so we have  
1374      to scan along to skip the extra bytes. All opcodes are less than 128, so we      code += _pcre_OP_lengths[c];
1375      can use relatively efficient code. */  
1376      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1377      a multi-byte character. The length in the table is a minimum, so we have to
1378      arrange to skip the extra bytes. */
1379    
1380    #ifdef SUPPORT_UTF8
1381      if (utf8) switch(c)      if (utf8) switch(c)
1382        {        {
1383        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1385  for (;;)
1385        case OP_EXACT:        case OP_EXACT:
1386        case OP_UPTO:        case OP_UPTO:
1387        case OP_MINUPTO:        case OP_MINUPTO:
1388          case OP_POSUPTO:
1389        case OP_STAR:        case OP_STAR:
1390        case OP_MINSTAR:        case OP_MINSTAR:
1391          case OP_POSSTAR:
1392        case OP_PLUS:        case OP_PLUS:
1393        case OP_MINPLUS:        case OP_MINPLUS:
1394          case OP_POSPLUS:
1395        case OP_QUERY:        case OP_QUERY:
1396        case OP_MINQUERY:        case OP_MINQUERY:
1397        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1398        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
1399        break;        break;
1400        }        }
1401  #endif  #endif
# Line 1072  Returns:      pointer to the opcode for Line 1422  Returns:      pointer to the opcode for
1422  static const uschar *  static const uschar *
1423  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1424  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1425  for (;;)  for (;;)
1426    {    {
1427    register int c = *code;    register int c = *code;
1428    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1429    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1430    else if (c > OP_BRA)  
1431      {    /* XCLASS is used for classes that cannot be represented just by a bit
1432      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1433      }    the table is zero; the actual length is stored in the compiled code. */
1434    
1435      if (c == OP_XCLASS) code += GET(code, 1);
1436    
1437      /* Otherwise, we can get the item's length from the table, except that for
1438      repeated character types, we have to test for \p and \P, which have an extra
1439      two bytes of parameters. */
1440    
1441    else    else
1442      {      {
1443      code += _pcre_OP_lengths[c];      switch(c)
1444          {
1445          case OP_TYPESTAR:
1446          case OP_TYPEMINSTAR:
1447          case OP_TYPEPLUS:
1448          case OP_TYPEMINPLUS:
1449          case OP_TYPEQUERY:
1450          case OP_TYPEMINQUERY:
1451          case OP_TYPEPOSSTAR:
1452          case OP_TYPEPOSPLUS:
1453          case OP_TYPEPOSQUERY:
1454          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1455          break;
1456    
1457  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1458          case OP_TYPEUPTO:
1459          case OP_TYPEMINUPTO:
1460          case OP_TYPEEXACT:
1461          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1462          break;
1463          }
1464    
1465        /* Add in the fixed length from the table */
1466    
1467        code += _pcre_OP_lengths[c];
1468    
1469      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
1470      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
1471      to scan along to skip the extra bytes. All opcodes are less than 128, so we      to arrange to skip the extra bytes. */
     can use relatively efficient code. */  
1472    
1473    #ifdef SUPPORT_UTF8
1474      if (utf8) switch(c)      if (utf8) switch(c)
1475        {        {
1476        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1478  for (;;)
1478        case OP_EXACT:        case OP_EXACT:
1479        case OP_UPTO:        case OP_UPTO:
1480        case OP_MINUPTO:        case OP_MINUPTO:
1481          case OP_POSUPTO:
1482        case OP_STAR:        case OP_STAR:
1483        case OP_MINSTAR:        case OP_MINSTAR:
1484          case OP_POSSTAR:
1485        case OP_PLUS:        case OP_PLUS:
1486        case OP_MINPLUS:        case OP_MINPLUS:
1487          case OP_POSPLUS:
1488        case OP_QUERY:        case OP_QUERY:
1489        case OP_MINQUERY:        case OP_MINQUERY:
1490        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1491        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
1492        break;        break;
1493        }        }
1494  #endif  #endif
# Line 1132  for (;;) Line 1503  for (;;)
1503  *************************************************/  *************************************************/
1504    
1505  /* This function scans through a branch of a compiled pattern to see whether it  /* This function scans through a branch of a compiled pattern to see whether it
1506  can match the empty string or not. It is called only from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1507  below. Note that first_significant_code() skips over assertions. If we hit an  below and from compile_branch() when checking for an unlimited repeat of a
1508  unclosed bracket, we return "empty" - this means we've struck an inner bracket  group that can match nothing. Note that first_significant_code() skips over
1509  whose current branch will already have been scanned.  assertions. If we hit an unclosed bracket, we return "empty" - this means we've
1510    struck an inner bracket whose current branch will already have been scanned.
1511    
1512  Arguments:  Arguments:
1513    code        points to start of search    code        points to start of search
# Line 1149  static BOOL Line 1521  static BOOL
1521  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1522  {  {
1523  register int c;  register int c;
1524  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
1525       code < endcode;       code < endcode;
1526       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1527    {    {
# Line 1157  for (code = first_significant_code(code Line 1529  for (code = first_significant_code(code
1529    
1530    c = *code;    c = *code;
1531    
1532    if (c >= OP_BRA)    /* Groups with zero repeats can of course be empty; skip them. */
1533    
1534      if (c == OP_BRAZERO || c == OP_BRAMINZERO)
1535        {
1536        code += _pcre_OP_lengths[c];
1537        do code += GET(code, 1); while (*code == OP_ALT);
1538        c = *code;
1539        continue;
1540        }
1541    
1542      /* For other groups, scan the branches. */
1543    
1544      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1545      {      {
1546      BOOL empty_branch;      BOOL empty_branch;
1547      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1173  for (code = first_significant_code(code Line 1557  for (code = first_significant_code(code
1557        }        }
1558      while (*code == OP_ALT);      while (*code == OP_ALT);
1559      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
     code += 1 + LINK_SIZE;  
1560      c = *code;      c = *code;
1561        continue;
1562      }      }
1563    
1564    else switch (c)    /* Handle the other opcodes */
1565    
1566      switch (c)
1567      {      {
1568      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1569        cannot be represented just by a bit map. This includes negated single
1570        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1571        actual length is stored in the compiled code, so we must update "code"
1572        here. */
1573    
1574  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1575      case OP_XCLASS:      case OP_XCLASS:
1576      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1577      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1578  #endif  #endif
1579    
# Line 1233  for (code = first_significant_code(code Line 1623  for (code = first_significant_code(code
1623      case OP_NOT:      case OP_NOT:
1624      case OP_PLUS:      case OP_PLUS:
1625      case OP_MINPLUS:      case OP_MINPLUS:
1626        case OP_POSPLUS:
1627      case OP_EXACT:      case OP_EXACT:
1628      case OP_NOTPLUS:      case OP_NOTPLUS:
1629      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1630        case OP_NOTPOSPLUS:
1631      case OP_NOTEXACT:      case OP_NOTEXACT:
1632      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1633      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1634        case OP_TYPEPOSPLUS:
1635      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1636      return FALSE;      return FALSE;
1637    
1638        /* These are going to continue, as they may be empty, but we have to
1639        fudge the length for the \p and \P cases. */
1640    
1641        case OP_TYPESTAR:
1642        case OP_TYPEMINSTAR:
1643        case OP_TYPEPOSSTAR:
1644        case OP_TYPEQUERY:
1645        case OP_TYPEMINQUERY:
1646        case OP_TYPEPOSQUERY:
1647        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1648        break;
1649    
1650        /* Same for these */
1651    
1652        case OP_TYPEUPTO:
1653        case OP_TYPEMINUPTO:
1654        case OP_TYPEPOSUPTO:
1655        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1656        break;
1657    
1658      /* End of branch */      /* End of branch */
1659    
1660      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 1663  for (code = first_significant_code(code
1663      case OP_ALT:      case OP_ALT:
1664      return TRUE;      return TRUE;
1665    
1666      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1667      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1668    
1669  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1670      case OP_STAR:      case OP_STAR:
1671      case OP_MINSTAR:      case OP_MINSTAR:
1672        case OP_POSSTAR:
1673      case OP_QUERY:      case OP_QUERY:
1674      case OP_MINQUERY:      case OP_MINQUERY:
1675        case OP_POSQUERY:
1676      case OP_UPTO:      case OP_UPTO:
1677      case OP_MINUPTO:      case OP_MINUPTO:
1678        case OP_POSUPTO:
1679      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1680      break;      break;
1681  #endif  #endif
# Line 1355  Returns:     a value representing the na Line 1771  Returns:     a value representing the na
1771  static int  static int
1772  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1773  {  {
1774    const char *pn = posix_names;
1775  register int yield = 0;  register int yield = 0;
1776  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1777    {    {
1778    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1779      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1780      pn += posix_name_lengths[yield] + 1;
1781    yield++;    yield++;
1782    }    }
1783  return -1;  return -1;
# Line 1377  earlier groups that are outside the curr Line 1795  earlier groups that are outside the curr
1795  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before
1796  it, after it has been compiled. This means that any OP_RECURSE items within it  it, after it has been compiled. This means that any OP_RECURSE items within it
1797  that refer to the group itself or any contained groups have to have their  that refer to the group itself or any contained groups have to have their
1798  offsets adjusted. That is the job of this function. Before it is called, the  offsets adjusted. That one of the jobs of this function. Before it is called,
1799  partially compiled regex must be temporarily terminated with OP_END.  the partially compiled regex must be temporarily terminated with OP_END.
1800    
1801    This function has been extended with the possibility of forward references for
1802    recursions and subroutine calls. It must also check the list of such references
1803    for the group we are dealing with. If it finds that one of the recursions in
1804    the current group is on this list, it adjusts the offset in the list, not the
1805    value in the reference (which is a group number).
1806    
1807  Arguments:  Arguments:
1808    group      points to the start of the group    group      points to the start of the group
1809    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1810    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1811    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1812      save_hwm   the hwm forward reference pointer at the start of the group
1813    
1814  Returns:     nothing  Returns:     nothing
1815  */  */
1816    
1817  static void  static void
1818  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1819      uschar *save_hwm)
1820  {  {
1821  uschar *ptr = group;  uschar *ptr = group;
1822    
1823  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1824    {    {
1825    int offset = GET(ptr, 1);    int offset;
1826    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1827    
1828      /* See if this recursion is on the forward reference list. If so, adjust the
1829      reference. */
1830    
1831      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1832        {
1833        offset = GET(hc, 0);
1834        if (cd->start_code + offset == ptr + 1)
1835          {
1836          PUT(hc, 0, offset + adjust);
1837          break;
1838          }
1839        }
1840    
1841      /* Otherwise, adjust the recursion offset if it's after the start of this
1842      group. */
1843    
1844      if (hc >= cd->hwm)
1845        {
1846        offset = GET(ptr, 1);
1847        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1848        }
1849    
1850    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1851    }    }
1852  }  }
# Line 1475  Yield:        TRUE when range returned; Line 1925  Yield:        TRUE when range returned;
1925  */  */
1926    
1927  static BOOL  static BOOL
1928  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
1929      unsigned int *odptr)
1930  {  {
1931  int c, chartype, othercase, next;  unsigned int c, othercase, next;
1932    
1933  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
1934    {    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
1935    
1936  if (c > d) return FALSE;  if (c > d) return FALSE;
1937    
# Line 1492  next = othercase + 1; Line 1940  next = othercase + 1;
1940    
1941  for (++c; c <= d; c++)  for (++c; c <= d; c++)
1942    {    {
1943    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (_pcre_ucp_othercase(c) != next) break;
         othercase != next)  
     break;  
1944    next++;    next++;
1945    }    }
1946    
# Line 1506  return TRUE; Line 1952  return TRUE;
1952  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
1953    
1954    
1955    
1956  /*************************************************  /*************************************************
1957  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
1958  *************************************************/  *************************************************/
1959    
1960  /* Scan the pattern, compiling it into the code vector. If the options are  /* This function is called for unlimited repeats of certain items, to see
1961  changed during the branch, the pointer is used to change the external options  whether the next thing could possibly match the repeated item. If not, it makes
1962  bits.  sense to automatically possessify the repeated item.
1963    
1964  Arguments:  Arguments:
1965    optionsptr     pointer to the option bits    op_code       the repeated op code
1966    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
1967    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
1968    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
1969    errorcodeptr   points to error code variable    ptr           next character in pattern
1970    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
1971    reqbyteptr     set to the last literal character required, else < 0    cd            contains pointers to tables etc.
   bcptr          points to current branch chain  
   cd             contains pointers to tables etc.  
1972    
1973  Returns:         TRUE on success  Returns:        TRUE if possessifying is wanted
                  FALSE, with *errorcodeptr set non-zero on error  
1974  */  */
1975    
1976  static BOOL  static BOOL
1977  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
1978    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const uschar *ptr, int options, compile_data *cd)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
1979  {  {
1980  int repeat_type, op_type;  int next;
1981  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  
1982  int bravalue = 0;  /* Skip whitespace and comments in extended mode */
1983  int greedy_default, greedy_non_default;  
1984  int firstbyte, reqbyte;  if ((options & PCRE_EXTENDED) != 0)
1985  int zeroreqbyte, zerofirstbyte;    {
1986  int req_caseopt, reqvary, tempreqvary;    for (;;)
1987  int condcount = 0;      {
1988  int options = *optionsptr;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
1989  int after_manual_callout = 0;      if (*ptr == '#')
1990  register int c;        {
1991  register uschar *code = *codeptr;        while (*(++ptr) != 0)
1992  uschar *tempcode;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
1993  BOOL inescq = FALSE;        }
1994        else break;
1995        }
1996      }
1997    
1998    /* If the next item is one that we can handle, get its value. A non-negative
1999    value is a character, a negative value is an escape value. */
2000    
2001    if (*ptr == '\\')
2002      {
2003      int temperrorcode = 0;
2004      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2005      if (temperrorcode != 0) return FALSE;
2006      ptr++;    /* Point after the escape sequence */
2007      }
2008    
2009    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2010      {
2011    #ifdef SUPPORT_UTF8
2012      if (utf8) { GETCHARINC(next, ptr); } else
2013    #endif
2014      next = *ptr++;
2015      }
2016    
2017    else return FALSE;
2018    
2019    /* Skip whitespace and comments in extended mode */
2020    
2021    if ((options & PCRE_EXTENDED) != 0)
2022      {
2023      for (;;)
2024        {
2025        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2026        if (*ptr == '#')
2027          {
2028          while (*(++ptr) != 0)
2029            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2030          }
2031        else break;
2032        }
2033      }
2034    
2035    /* If the next thing is itself optional, we have to give up. */
2036    
2037    if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
2038      return FALSE;
2039    
2040    /* Now compare the next item with the previous opcode. If the previous is a
2041    positive single character match, "item" either contains the character or, if
2042    "item" is greater than 127 in utf8 mode, the character's bytes are in
2043    utf8_char. */
2044    
2045    
2046    /* Handle cases when the next item is a character. */
2047    
2048    if (next >= 0) switch(op_code)
2049      {
2050      case OP_CHAR:
2051    #ifdef SUPPORT_UTF8
2052      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2053    #endif
2054      return item != next;
2055    
2056      /* For CHARNC (caseless character) we must check the other case. If we have
2057      Unicode property support, we can use it to test the other case of
2058      high-valued characters. */
2059    
2060      case OP_CHARNC:
2061    #ifdef SUPPORT_UTF8
2062      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2063    #endif
2064      if (item == next) return FALSE;
2065    #ifdef SUPPORT_UTF8
2066      if (utf8)
2067        {
2068        unsigned int othercase;
2069        if (next < 128) othercase = cd->fcc[next]; else
2070    #ifdef SUPPORT_UCP
2071        othercase = _pcre_ucp_othercase((unsigned int)next);
2072    #else
2073        othercase = NOTACHAR;
2074    #endif
2075        return (unsigned int)item != othercase;
2076        }
2077      else
2078    #endif  /* SUPPORT_UTF8 */
2079      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2080    
2081      /* For OP_NOT, "item" must be a single-byte character. */
2082    
2083      case OP_NOT:
2084      if (next < 0) return FALSE;  /* Not a character */
2085      if (item == next) return TRUE;
2086      if ((options & PCRE_CASELESS) == 0) return FALSE;
2087    #ifdef SUPPORT_UTF8
2088      if (utf8)
2089        {
2090        unsigned int othercase;
2091        if (next < 128) othercase = cd->fcc[next]; else
2092    #ifdef SUPPORT_UCP
2093        othercase = _pcre_ucp_othercase(next);
2094    #else
2095        othercase = NOTACHAR;
2096    #endif
2097        return (unsigned int)item == othercase;
2098        }
2099      else
2100    #endif  /* SUPPORT_UTF8 */
2101      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2102    
2103      case OP_DIGIT:
2104      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2105    
2106      case OP_NOT_DIGIT:
2107      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2108    
2109      case OP_WHITESPACE:
2110      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2111    
2112      case OP_NOT_WHITESPACE:
2113      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2114    
2115      case OP_WORDCHAR:
2116      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2117    
2118      case OP_NOT_WORDCHAR:
2119      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2120    
2121      case OP_HSPACE:
2122      case OP_NOT_HSPACE:
2123      switch(next)
2124        {
2125        case 0x09:
2126        case 0x20:
2127        case 0xa0:
2128        case 0x1680:
2129        case 0x180e:
2130        case 0x2000:
2131        case 0x2001:
2132        case 0x2002:
2133        case 0x2003:
2134        case 0x2004:
2135        case 0x2005:
2136        case 0x2006:
2137        case 0x2007:
2138        case 0x2008:
2139        case 0x2009:
2140        case 0x200A:
2141        case 0x202f:
2142        case 0x205f:
2143        case 0x3000:
2144        return op_code != OP_HSPACE;
2145        default:
2146        return op_code == OP_HSPACE;
2147        }
2148    
2149      case OP_VSPACE:
2150      case OP_NOT_VSPACE:
2151      switch(next)
2152        {
2153        case 0x0a:
2154        case 0x0b:
2155        case 0x0c:
2156        case 0x0d:
2157        case 0x85:
2158        case 0x2028:
2159        case 0x2029:
2160        return op_code != OP_VSPACE;
2161        default:
2162        return op_code == OP_VSPACE;
2163        }
2164    
2165      default:
2166      return FALSE;
2167      }
2168    
2169    
2170    /* Handle the case when the next item is \d, \s, etc. */
2171    
2172    switch(op_code)
2173      {
2174      case OP_CHAR:
2175      case OP_CHARNC:
2176    #ifdef SUPPORT_UTF8
2177      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2178    #endif
2179      switch(-next)
2180        {
2181        case ESC_d:
2182        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2183    
2184        case ESC_D:
2185        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2186    
2187        case ESC_s:
2188        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2189    
2190        case ESC_S:
2191        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2192    
2193        case ESC_w:
2194        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2195    
2196        case ESC_W:
2197        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2198    
2199        case ESC_h:
2200        case ESC_H:
2201        switch(item)
2202          {
2203          case 0x09:
2204          case 0x20:
2205          case 0xa0:
2206          case 0x1680:
2207          case 0x180e:
2208          case 0x2000:
2209          case 0x2001:
2210          case 0x2002:
2211          case 0x2003:
2212          case 0x2004:
2213          case 0x2005:
2214          case 0x2006:
2215          case 0x2007:
2216          case 0x2008:
2217          case 0x2009:
2218          case 0x200A:
2219          case 0x202f:
2220          case 0x205f:
2221          case 0x3000:
2222          return -next != ESC_h;
2223          default:
2224          return -next == ESC_h;
2225          }
2226    
2227        case ESC_v:
2228        case ESC_V:
2229        switch(item)
2230          {
2231          case 0x0a:
2232          case 0x0b:
2233          case 0x0c:
2234          case 0x0d:
2235          case 0x85:
2236          case 0x2028:
2237          case 0x2029:
2238          return -next != ESC_v;
2239          default:
2240          return -next == ESC_v;
2241          }
2242    
2243        default:
2244        return FALSE;
2245        }
2246    
2247      case OP_DIGIT:
2248      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2249             next == -ESC_h || next == -ESC_v;
2250    
2251      case OP_NOT_DIGIT:
2252      return next == -ESC_d;
2253    
2254      case OP_WHITESPACE:
2255      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2256    
2257      case OP_NOT_WHITESPACE:
2258      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2259    
2260      case OP_HSPACE:
2261      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2262    
2263      case OP_NOT_HSPACE:
2264      return next == -ESC_h;
2265    
2266      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2267      case OP_VSPACE:
2268      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2269    
2270      case OP_NOT_VSPACE:
2271      return next == -ESC_v;
2272    
2273      case OP_WORDCHAR:
2274      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2275    
2276      case OP_NOT_WORDCHAR:
2277      return next == -ESC_w || next == -ESC_d;
2278    
2279      default:
2280      return FALSE;
2281      }
2282    
2283    /* Control does not reach here */
2284    }
2285    
2286    
2287    
2288    /*************************************************
2289    *           Compile one branch                   *
2290    *************************************************/
2291    
2292    /* Scan the pattern, compiling it into the a vector. If the options are
2293    changed during the branch, the pointer is used to change the external options
2294    bits. This function is used during the pre-compile phase when we are trying
2295    to find out the amount of memory needed, as well as during the real compile
2296    phase. The value of lengthptr distinguishes the two phases.
2297    
2298    Arguments:
2299      optionsptr     pointer to the option bits
2300      codeptr        points to the pointer to the current code point
2301      ptrptr         points to the current pattern pointer
2302      errorcodeptr   points to error code variable
2303      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2304      reqbyteptr     set to the last literal character required, else < 0
2305      bcptr          points to current branch chain
2306      cd             contains pointers to tables etc.
2307      lengthptr      NULL during the real compile phase
2308                     points to length accumulator during pre-compile phase
2309    
2310    Returns:         TRUE on success
2311                     FALSE, with *errorcodeptr set non-zero on error
2312    */
2313    
2314    static BOOL
2315    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2316      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2317      compile_data *cd, int *lengthptr)
2318    {
2319    int repeat_type, op_type;
2320    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2321    int bravalue = 0;
2322    int greedy_default, greedy_non_default;
2323    int firstbyte, reqbyte;
2324    int zeroreqbyte, zerofirstbyte;
2325    int req_caseopt, reqvary, tempreqvary;
2326    int options = *optionsptr;
2327    int after_manual_callout = 0;
2328    int length_prevgroup = 0;
2329    register int c;
2330    register uschar *code = *codeptr;
2331    uschar *last_code = code;
2332    uschar *orig_code = code;
2333    uschar *tempcode;
2334    BOOL inescq = FALSE;
2335  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2336  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2337  const uschar *tempptr;  const uschar *tempptr;
2338  uschar *previous = NULL;  uschar *previous = NULL;
2339  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2340    uschar *save_hwm = NULL;
2341  uschar classbits[32];  uschar classbits[32];
2342    
2343  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1562  uschar *class_utf8data; Line 2347  uschar *class_utf8data;
2347  uschar utf8_char[6];  uschar utf8_char[6];
2348  #else  #else
2349  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2350    uschar *utf8_char = NULL;
2351    #endif
2352    
2353    #ifdef DEBUG
2354    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2355  #endif  #endif
2356    
2357  /* Set up the default and non-default settings for greediness */  /* Set up the default and non-default settings for greediness */
# Line 1595  for (;; ptr++) Line 2385  for (;; ptr++)
2385    BOOL negate_class;    BOOL negate_class;
2386    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2387    BOOL is_quantifier;    BOOL is_quantifier;
2388      BOOL is_recurse;
2389      BOOL reset_bracount;
2390    int class_charcount;    int class_charcount;
2391    int class_lastchar;    int class_lastchar;
2392    int newoptions;    int newoptions;
2393    int recno;    int recno;
2394      int refsign;
2395    int skipbytes;    int skipbytes;
2396    int subreqbyte;    int subreqbyte;
2397    int subfirstbyte;    int subfirstbyte;
2398      int terminator;
2399    int mclength;    int mclength;
2400    uschar mcbuffer[8];    uschar mcbuffer[8];
2401    
2402    /* Next byte in the pattern */    /* Get next byte in the pattern */
2403    
2404    c = *ptr;    c = *ptr;
2405    
2406      /* If we are in the pre-compile phase, accumulate the length used for the
2407      previous cycle of this loop. */
2408    
2409      if (lengthptr != NULL)
2410        {
2411    #ifdef DEBUG
2412        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2413    #endif
2414        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2415          {
2416          *errorcodeptr = ERR52;
2417          goto FAILED;
2418          }
2419    
2420        /* There is at least one situation where code goes backwards: this is the
2421        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2422        the class is simply eliminated. However, it is created first, so we have to
2423        allow memory for it. Therefore, don't ever reduce the length at this point.
2424        */
2425    
2426        if (code < last_code) code = last_code;
2427    
2428        /* Paranoid check for integer overflow */
2429    
2430        if (OFLOW_MAX - *lengthptr < code - last_code)
2431          {
2432          *errorcodeptr = ERR20;
2433          goto FAILED;
2434          }
2435    
2436        *lengthptr += code - last_code;
2437        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2438    
2439        /* If "previous" is set and it is not at the start of the work space, move
2440        it back to there, in order to avoid filling up the work space. Otherwise,
2441        if "previous" is NULL, reset the current code pointer to the start. */
2442    
2443        if (previous != NULL)
2444          {
2445          if (previous > orig_code)
2446            {
2447            memmove(orig_code, previous, code - previous);
2448            code -= previous - orig_code;
2449            previous = orig_code;
2450            }
2451          }
2452        else code = orig_code;
2453    
2454        /* Remember where this code item starts so we can pick up the length
2455        next time round. */
2456    
2457        last_code = code;
2458        }
2459    
2460      /* In the real compile phase, just check the workspace used by the forward
2461      reference list. */
2462    
2463      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2464        {
2465        *errorcodeptr = ERR52;
2466        goto FAILED;
2467        }
2468    
2469    /* 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 */
2470    
2471    if (inescq && c != 0)    if (inescq && c != 0)
# Line 1623  for (;; ptr++) Line 2480  for (;; ptr++)
2480        {        {
2481        if (previous_callout != NULL)        if (previous_callout != NULL)
2482          {          {
2483          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2484              complete_callout(previous_callout, ptr, cd);
2485          previous_callout = NULL;          previous_callout = NULL;
2486          }          }
2487        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1644  for (;; ptr++) Line 2502  for (;; ptr++)
2502    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2503         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2504      {      {
2505      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2506          complete_callout(previous_callout, ptr, cd);
2507      previous_callout = NULL;      previous_callout = NULL;
2508      }      }
2509    
# Line 1655  for (;; ptr++) Line 2514  for (;; ptr++)
2514      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2515      if (c == '#')      if (c == '#')
2516        {        {
2517        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2518        on the Macintosh. */          {
2519        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2520        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2521          if (*ptr != 0) continue;
2522    
2523          /* Else fall through to handle end of string */
2524          c = 0;
2525        }        }
2526      }      }
2527    
# Line 1672  for (;; ptr++) Line 2535  for (;; ptr++)
2535    
2536    switch(c)    switch(c)
2537      {      {
2538      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2539        case 0:                        /* The branch terminates at string end */
2540      case 0:      case '|':                      /* or | or ) */
     case '|':  
2541      case ')':      case ')':
2542      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2543      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2544      *codeptr = code;      *codeptr = code;
2545      *ptrptr = ptr;      *ptrptr = ptr;
2546        if (lengthptr != NULL)
2547          {
2548          if (OFLOW_MAX - *lengthptr < code - last_code)
2549            {
2550            *errorcodeptr = ERR20;
2551            goto FAILED;
2552            }
2553          *lengthptr += code - last_code;   /* To include callout length */
2554          DPRINTF((">> end branch\n"));
2555          }
2556      return TRUE;      return TRUE;
2557    
2558    
2559        /* ===================================================================*/
2560      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2561      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2562    
# Line 1711  for (;; ptr++) Line 2585  for (;; ptr++)
2585      *code++ = OP_ANY;      *code++ = OP_ANY;
2586      break;      break;
2587    
2588      /* Character classes. If the included characters are all < 255 in value, we  
2589      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2590      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
2591      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
2592      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2593        map as usual, then invert it at the end. However, we use a different opcode
2594        so that data characters > 255 can be handled correctly.
2595    
2596      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2597      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
# Line 1736  for (;; ptr++) Line 2612  for (;; ptr++)
2612        goto FAILED;        goto FAILED;
2613        }        }
2614    
2615      /* 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,
2616        if the first few characters (either before or after ^) are \Q\E or \E we
2617        skip them too. This makes for compatibility with Perl. */
2618    
2619      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2620        for (;;)
2621        {        {
       negate_class = TRUE;  
2622        c = *(++ptr);        c = *(++ptr);
2623        }        if (c == '\\')
2624      else          {
2625        {          if (ptr[1] == 'E') ptr++;
2626        negate_class = FALSE;            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2627                else break;
2628            }
2629          else if (!negate_class && c == '^')
2630            negate_class = TRUE;
2631          else break;
2632        }        }
2633    
2634      /* 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
2635      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
2636      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2637    
2638      class_charcount = 0;      class_charcount = 0;
2639      class_lastchar = -1;      class_lastchar = -1;
2640    
2641        /* Initialize the 32-char bit map to all zeros. We build the map in a
2642        temporary bit of memory, in case the class contains only 1 character (less
2643        than 256), because in that case the compiled code doesn't use the bit map.
2644        */
2645    
2646        memset(classbits, 0, 32 * sizeof(uschar));
2647    
2648  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2649      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2650      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2651  #endif  #endif
2652    
     /* 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));  
   
2653      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2654      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
2655      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. */  
2656    
2657      do      if (c != 0) do
2658        {        {
2659          const uschar *oldptr;
2660    
2661  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2662        if (utf8 && c > 127)        if (utf8 && c > 127)
2663          {                           /* Braces are required because the */          {                           /* Braces are required because the */
# Line 1786  for (;; ptr++) Line 2669  for (;; ptr++)
2669    
2670        if (inescq)        if (inescq)
2671          {          {
2672          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2673            {            {
2674            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2675            ptr++;            ptr++;                            /* Skip the 'E' */
2676            continue;            continue;                         /* Carry on with next */
2677            }            }
2678          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2679          }          }
2680    
2681        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1806  for (;; ptr++) Line 2689  for (;; ptr++)
2689            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr, cd))
2690          {          {
2691          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2692          int posix_class, i;          int posix_class, taboffset, tabopt;
2693          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2694            uschar pbits[32];
2695    
2696          if (ptr[1] != ':')          if (ptr[1] != ':')
2697            {            {
# Line 1836  for (;; ptr++) Line 2720  for (;; ptr++)
2720          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
2721            posix_class = 0;            posix_class = 0;
2722    
2723          /* 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
2724          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
2725          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
2726          white space chars afterwards. */          result into the bit map that is being built. */
2727    
2728          posix_class *= 3;          posix_class *= 3;
2729          for (i = 0; i < 3; i++)  
2730            /* Copy in the first table (always present) */
2731    
2732            memcpy(pbits, cbits + posix_class_maps[posix_class],
2733              32 * sizeof(uschar));
2734    
2735            /* If there is a second table, add or remove it as required. */
2736    
2737            taboffset = posix_class_maps[posix_class + 1];
2738            tabopt = posix_class_maps[posix_class + 2];
2739    
2740            if (taboffset >= 0)
2741            {            {
2742            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
2743            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;  
             }  
2744            else            else
2745              {              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;  
             }  
2746            }            }
2747    
2748            /* Not see if we need to remove any special characters. An option
2749            value of 1 removes vertical space and 2 removes underscore. */
2750    
2751            if (tabopt < 0) tabopt = -tabopt;
2752            if (tabopt == 1) pbits[1] &= ~0x3c;
2753              else if (tabopt == 2) pbits[11] &= 0x7f;
2754    
2755            /* Add the POSIX table or its complement into the main table that is
2756            being built and we are done. */
2757    
2758            if (local_negate)
2759              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
2760            else
2761              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
2762    
2763          ptr = tempptr + 1;          ptr = tempptr + 1;
2764          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
2765          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
2766          }          }
2767    
2768        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2769        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
2770        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.
2771        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2772        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  
2773        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2774    
2775        if (c == '\\')        if (c == '\\')
2776          {          {
2777          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2778            if (*errorcodeptr != 0) goto FAILED;
2779    
2780          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */
2781          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 */
2782            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2783          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2784            {            {
2785            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1890  for (;; ptr++) Line 2789  for (;; ptr++)
2789            else inescq = TRUE;            else inescq = TRUE;
2790            continue;            continue;
2791            }            }
2792            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2793    
2794          if (c < 0)          if (c < 0)
2795            {            {
2796            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2797            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2798            switch (-c)  
2799              /* Save time by not doing this in the pre-compile phase. */
2800    
2801              if (lengthptr == NULL) switch (-c)
2802              {              {
2803              case ESC_d:              case ESC_d:
2804              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
# Line 1923  for (;; ptr++) Line 2826  for (;; ptr++)
2826              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2827              continue;              continue;
2828    
2829  #ifdef SUPPORT_UCP              case ESC_E: /* Perl ignores an orphan \E */
2830              case ESC_p:              continue;
2831              case ESC_P:  
2832                default:    /* Not recognized; fall through */
2833                break;      /* Need "default" setting to stop compiler warning. */
2834                }
2835    
2836              /* In the pre-compile phase, just do the recognition. */
2837    
2838              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2839                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
2840    
2841              /* We need to deal with \H, \h, \V, and \v in both phases because
2842              they use extra memory. */
2843    
2844              if (-c == ESC_h)
2845                {
2846                SETBIT(classbits, 0x09); /* VT */
2847                SETBIT(classbits, 0x20); /* SPACE */
2848                SETBIT(classbits, 0xa0); /* NSBP */
2849    #ifdef SUPPORT_UTF8
2850                if (utf8)
2851                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
2852                class_utf8 = TRUE;                class_utf8 = TRUE;
2853                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_SINGLE;
2854                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2855                *class_utf8data++ = property;                *class_utf8data++ = XCL_SINGLE;
2856                class_charcount -= 2;   /* Not a < 256 character */                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2857                  *class_utf8data++ = XCL_RANGE;
2858                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2859                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2860                  *class_utf8data++ = XCL_SINGLE;
2861                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2862                  *class_utf8data++ = XCL_SINGLE;
2863                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2864                  *class_utf8data++ = XCL_SINGLE;
2865                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2866                }                }
             continue;  
2867  #endif  #endif
2868                continue;
2869                }
2870    
2871              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_H)
2872              strict mode. By default, for compatibility with Perl, they are              {
2873              treated as literals. */              for (c = 0; c < 32; c++)
2874                  {
2875                  int x = 0xff;
2876                  switch (c)
2877                    {
2878                    case 0x09/8: x ^= 1 << (0x09%8); break;
2879                    case 0x20/8: x ^= 1 << (0x20%8); break;
2880                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2881                    default: break;
2882                    }
2883                  classbits[c] |= x;
2884                  }
2885    
2886              default:  #ifdef SUPPORT_UTF8
2887              if ((options & PCRE_EXTRA) != 0)              if (utf8)
2888                {                {
2889                *errorcodeptr = ERR7;                class_utf8 = TRUE;
2890                goto FAILED;                *class_utf8data++ = XCL_RANGE;
2891                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2892                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
2893                  *class_utf8data++ = XCL_RANGE;
2894                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
2895                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
2896                  *class_utf8data++ = XCL_RANGE;
2897                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
2898                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
2899                  *class_utf8data++ = XCL_RANGE;
2900                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
2901                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
2902                  *class_utf8data++ = XCL_RANGE;
2903                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
2904                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
2905                  *class_utf8data++ = XCL_RANGE;
2906                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
2907                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
2908                  *class_utf8data++ = XCL_RANGE;
2909                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
2910                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2911                }                }
2912              c = *ptr;              /* The final character */  #endif
2913              class_charcount -= 2;  /* Undo the default count from above */              continue;
2914              }              }
           }  
2915    
2916          /* Fall through if we have a single character (c >= 0). This may be            if (-c == ESC_v)
2917          > 256 in UTF-8 mode. */              {
2918                SETBIT(classbits, 0x0a); /* LF */
2919                SETBIT(classbits, 0x0b); /* VT */
2920                SETBIT(classbits, 0x0c); /* FF */
2921                SETBIT(classbits, 0x0d); /* CR */
2922                SETBIT(classbits, 0x85); /* NEL */
2923    #ifdef SUPPORT_UTF8
2924                if (utf8)
2925                  {
2926                  class_utf8 = TRUE;
2927                  *class_utf8data++ = XCL_RANGE;
2928                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
2929                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2930                  }
2931    #endif
2932                continue;
2933                }
2934    
2935              if (-c == ESC_V)
2936                {
2937                for (c = 0; c < 32; c++)
2938                  {
2939                  int x = 0xff;
2940                  switch (c)
2941                    {
2942                    case 0x0a/8: x ^= 1 << (0x0a%8);
2943                                 x ^= 1 << (0x0b%8);
2944                                 x ^= 1 << (0x0c%8);
2945                                 x ^= 1 << (0x0d%8);
2946                                 break;
2947                    case 0x85/8: x ^= 1 << (0x85%8); break;
2948                    default: break;
2949                    }
2950                  classbits[c] |= x;
2951                  }
2952    
2953    #ifdef SUPPORT_UTF8
2954                if (utf8)
2955                  {
2956                  class_utf8 = TRUE;
2957                  *class_utf8data++ = XCL_RANGE;
2958                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2959                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
2960                  *class_utf8data++ = XCL_RANGE;
2961                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2962                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2963                  }
2964    #endif
2965                continue;
2966                }
2967    
2968              /* We need to deal with \P and \p in both phases. */
2969    
2970    #ifdef SUPPORT_UCP
2971              if (-c == ESC_p || -c == ESC_P)
2972                {
2973                BOOL negated;
2974                int pdata;
2975                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
2976                if (ptype < 0) goto FAILED;
2977                class_utf8 = TRUE;
2978                *class_utf8data++ = ((-c == ESC_p) != negated)?
2979                  XCL_PROP : XCL_NOTPROP;
2980                *class_utf8data++ = ptype;
2981                *class_utf8data++ = pdata;
2982                class_charcount -= 2;   /* Not a < 256 character */
2983                continue;
2984                }
2985    #endif
2986              /* Unrecognized escapes are faulted if PCRE is running in its
2987              strict mode. By default, for compatibility with Perl, they are
2988              treated as literals. */
2989    
2990              if ((options & PCRE_EXTRA) != 0)
2991                {
2992                *errorcodeptr = ERR7;
2993                goto FAILED;
2994                }
2995    
2996              class_charcount -= 2;  /* Undo the default count from above */
2997              c = *ptr;              /* Get the final character and fall through */
2998              }
2999    
3000            /* Fall through if we have a single character (c >= 0). This may be
3001            greater than 256 in UTF-8 mode. */
3002    
3003          }   /* End of backslash handling */          }   /* End of backslash handling */
3004    
3005        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3006        Perl does not permit ']' to be the end of the range. A '-' character        Perl does not permit ']' to be the end of the range. A '-' character
3007        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3008          entirely. The code for handling \Q and \E is messy. */
3009    
3010          CHECK_RANGE:
3011          while (ptr[1] == '\\' && ptr[2] == 'E')
3012            {
3013            inescq = FALSE;
3014            ptr += 2;
3015            }
3016    
3017          oldptr = ptr;
3018    
3019          /* Remember \r or \n */
3020    
3021          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3022    
3023        if (ptr[1] == '-' && ptr[2] != ']')        /* Check for range */
3024    
3025          if (!inescq && ptr[1] == '-')
3026          {          {
3027          int d;          int d;
3028          ptr += 2;          ptr += 2;
3029            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
3030    
3031            /* If we hit \Q (not followed by \E) at this point, go into escaped
3032            mode. */
3033    
3034            while (*ptr == '\\' && ptr[1] == 'Q')
3035              {
3036              ptr += 2;
3037              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
3038              inescq = TRUE;
3039              break;
3040              }
3041    
3042            if (*ptr == 0 || (!inescq && *ptr == ']'))
3043              {
3044              ptr = oldptr;
3045              goto LONE_SINGLE_CHARACTER;
3046              }
3047    
3048  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3049          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3058  for (;; ptr++)
3058          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
3059          in such circumstances. */          in such circumstances. */
3060    
3061          if (d == '\\')          if (!inescq && d == '\\')
3062            {            {
3063            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3064            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3065    
3066            /* \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
3067            was literal */            special means the '-' was literal */
3068    
3069            if (d < 0)            if (d < 0)
3070              {              {
3071              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = '\b';
3072              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = 'X';
3073                else if (d == -ESC_R) d = 'R'; else
3074                {                {
3075                ptr = oldptr - 2;                ptr = oldptr;
3076                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3077                }                }
3078              }              }
3079            }            }
3080    
3081          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3082          the pre-pass. Optimize one-character ranges */          one-character ranges */
3083    
3084            if (d < c)
3085              {
3086              *errorcodeptr = ERR8;
3087              goto FAILED;
3088              }
3089    
3090          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3091    
3092            /* Remember \r or \n */
3093    
3094            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3095    
3096          /* 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
3097          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3098          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2022  for (;; ptr++) Line 3110  for (;; ptr++)
3110  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3111            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3112              {              {
3113              int occ, ocd;              unsigned int occ, ocd;
3114              int cc = c;              unsigned int cc = c;
3115              int origd = d;              unsigned int origd = d;
3116              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3117                {                {
3118                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3119                      ocd <= (unsigned int)d)
3120                    continue;                          /* Skip embedded ranges */
3121    
3122                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3123                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3124                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3125                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3126                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3127                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3128                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3129                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3130                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3131                  d = ocd;                  d = ocd;
3132                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3174  for (;; ptr++)
3174          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
3175          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3176    
3177          for (; c <= d; c++)          class_charcount += d - c + 1;
3178            class_lastchar = d;
3179    
3180            /* We can save a bit of time by skipping this in the pre-compile. */
3181    
3182            if (lengthptr == NULL) for (; c <= d; c++)
3183            {            {
3184            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3185            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3187  for (;; ptr++)
3187              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3188              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3189              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3190            }            }
3191    
3192          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3210  for (;; ptr++)
3210  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3211          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3212            {            {
3213            int chartype;            unsigned int othercase;
3214            int othercase;            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3215              {              {
3216              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3217              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3236  for (;; ptr++)
3236          }          }
3237        }        }
3238    
3239      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3240      loop. This "while" is the end of the "do" above. */  
3241        while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
3242    
3243        if (c == 0)                          /* Missing terminating ']' */
3244          {
3245          *errorcodeptr = ERR6;
3246          goto FAILED;
3247          }
3248    
3249    
3250    /* This code has been disabled because it would mean that \s counts as
3251    an explicit \r or \n reference, and that's not really what is wanted. Now
3252    we set the flag only if there is a literal "\r" or "\n" in the class. */
3253    
3254    #if 0
3255        /* Remember whether \r or \n are in this class */
3256    
3257        if (negate_class)
3258          {
3259          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3260          }
3261        else
3262          {
3263          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3264          }
3265    #endif
3266    
     while ((c = *(++ptr)) != ']' || inescq);  
3267    
3268      /* 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
3269      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
3270      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
3271      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3272      single-bytes only. This is an historical hangover. Maybe one day we can  
3273      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3274        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3275        operate on single-bytes only. This is an historical hangover. Maybe one day
3276        we can tidy these opcodes to handle multi-byte characters.
3277    
3278      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
3279      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2163  for (;; ptr++) Line 3283  for (;; ptr++)
3283      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3284    
3285  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3286      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3287            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3288  #else  #else
3289      if (class_charcount == 1)      if (class_charcount == 1)
3290  #endif  #endif
# Line 2210  for (;; ptr++) Line 3328  for (;; ptr++)
3328    
3329      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3330      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode. If there are no characters < 256,
3331      we can omit the bitmap. */      we can omit the bitmap in the actual compiled code. */
3332    
3333  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3334      if (class_utf8)      if (class_utf8)
# Line 2220  for (;; ptr++) Line 3338  for (;; ptr++)
3338        code += LINK_SIZE;        code += LINK_SIZE;
3339        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3340    
3341        /* 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;
3342        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3343    
3344        if (class_charcount > 0)        if (class_charcount > 0)
3345          {          {
3346          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3347            memmove(code + 32, code, class_utf8data - code);
3348          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3349          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;  
3350          }          }
3351          else code = class_utf8data;
3352    
3353        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3354    
# Line 2254  for (;; ptr++) Line 3365  for (;; ptr++)
3365      if (negate_class)      if (negate_class)
3366        {        {
3367        *code++ = OP_NCLASS;        *code++ = OP_NCLASS;
3368        for (c = 0; c < 32; c++) code[c] = ~classbits[c];        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3369            for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3370        }        }
3371      else      else
3372        {        {
# Line 2264  for (;; ptr++) Line 3376  for (;; ptr++)
3376      code += 32;      code += 32;
3377      break;      break;
3378    
3379    
3380        /* ===================================================================*/
3381      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3382      has been tested above. */      has been tested above. */
3383    
# Line 2331  for (;; ptr++) Line 3445  for (;; ptr++)
3445        }        }
3446      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3447    
     /* 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;  
       }  
   
3448      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3449      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
3450      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
# Line 2378  for (;; ptr++) Line 3478  for (;; ptr++)
3478          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3479          }          }
3480    
3481          /* If the repetition is unlimited, it pays to see if the next thing on
3482          the line is something that cannot possibly match this character. If so,
3483          automatically possessifying this item gains some performance in the case
3484          where the match fails. */
3485    
3486          if (!possessive_quantifier &&
3487              repeat_max < 0 &&
3488              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3489                options, cd))
3490            {
3491            repeat_type = 0;    /* Force greedy */
3492            possessive_quantifier = TRUE;
3493            }
3494    
3495        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3496        }        }
3497    
3498      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3499      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-
3500      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3501      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3502        currently used only for single-byte chars. */
3503    
3504      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3505        {        {
3506        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3507        c = previous[1];        c = previous[1];
3508          if (!possessive_quantifier &&
3509              repeat_max < 0 &&
3510              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3511            {
3512            repeat_type = 0;    /* Force greedy */
3513            possessive_quantifier = TRUE;
3514            }
3515        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3516        }        }
3517    
# Line 2403  for (;; ptr++) Line 3525  for (;; ptr++)
3525      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3526        {        {
3527        uschar *oldcode;        uschar *oldcode;
3528        int prop_type;        int prop_type, prop_value;
3529        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3530        c = *previous;        c = *previous;
3531    
3532          if (!possessive_quantifier &&
3533              repeat_max < 0 &&
3534              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3535            {
3536            repeat_type = 0;    /* Force greedy */
3537            possessive_quantifier = TRUE;
3538            }
3539    
3540        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3541        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3542          previous[1] : -1;          {
3543            prop_type = previous[1];
3544            prop_value = previous[2];
3545            }
3546          else prop_type = prop_value = -1;
3547    
3548        oldcode = code;        oldcode = code;
3549        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2422  for (;; ptr++) Line 3556  for (;; ptr++)
3556        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3557        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3558    
3559        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3560    
3561        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3562    
# Line 2443  for (;; ptr++) Line 3577  for (;; ptr++)
3577          }          }
3578    
3579        /* 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
3580        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3581        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
3582        one less than the maximum. */        one less than the maximum. */
3583    
# Line 2470  for (;; ptr++) Line 3604  for (;; ptr++)
3604    
3605          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3606          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
3607          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3608          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3609          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3610    
# Line 2486  for (;; ptr++) Line 3620  for (;; ptr++)
3620  #endif  #endif
3621              {              {
3622              *code++ = c;              *code++ = c;
3623              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3624                  {
3625                  *code++ = prop_type;
3626                  *code++ = prop_value;
3627                  }
3628              }              }
3629            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3630            }            }
3631    
3632          /* 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
3633          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3634            UPTO is just for 1 instance, we can use QUERY instead. */
3635    
3636          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3637            {            {
# Line 2505  for (;; ptr++) Line 3644  for (;; ptr++)
3644            else            else
3645  #endif  #endif
3646            *code++ = c;            *code++ = c;
3647            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3648                {
3649                *code++ = prop_type;
3650                *code++ = prop_value;
3651                }
3652            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3653            *code++ = OP_UPTO + repeat_type;  
3654            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3655                {
3656                *code++ = OP_QUERY + repeat_type;
3657                }
3658              else
3659                {
3660                *code++ = OP_UPTO + repeat_type;
3661                PUT2INC(code, 0, repeat_max);
3662                }
3663            }            }
3664          }          }
3665    
# Line 2524  for (;; ptr++) Line 3675  for (;; ptr++)
3675  #endif  #endif
3676        *code++ = c;        *code++ = c;
3677    
3678        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3679        defines the required property. */        define the required property. */
3680    
3681  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3682        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3683            {
3684            *code++ = prop_type;
3685            *code++ = prop_value;
3686            }
3687  #endif  #endif
3688        }        }
3689    
# Line 2551  for (;; ptr++) Line 3706  for (;; ptr++)
3706        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3707        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3708    
3709        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3710    
3711        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3712          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2571  for (;; ptr++) Line 3726  for (;; ptr++)
3726      /* 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
3727      cases. */      cases. */
3728    
3729      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3730               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3731        {        {
3732        register int i;        register int i;
3733        int ketoffset = 0;        int ketoffset = 0;
3734        int len = code - previous;        int len = code - previous;
3735        uschar *bralink = NULL;        uschar *bralink = NULL;
3736    
3737          /* Repeating a DEFINE group is pointless */
3738    
3739          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3740            {
3741            *errorcodeptr = ERR55;
3742            goto FAILED;
3743            }
3744    
3745        /* 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
3746        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
3747        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 2613  for (;; ptr++) Line 3776  for (;; ptr++)
3776          /* 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
3777          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
3778          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
3779          any internal group, because the offset is from the start of the whole          any internal or forward referenced group, because the offset is from
3780          regex. Temporarily terminate the pattern while doing this. */          the start of the whole regex. Temporarily terminate the pattern while
3781            doing this. */
3782    
3783          if (repeat_max <= 1)          if (repeat_max <= 1)
3784            {            {
3785            *code = OP_END;            *code = OP_END;
3786            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3787            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3788            code++;            code++;
3789            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2637  for (;; ptr++) Line 3801  for (;; ptr++)
3801            {            {
3802            int offset;            int offset;
3803            *code = OP_END;            *code = OP_END;
3804            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3805            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3806            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3807            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 3821  for (;; ptr++)
3821        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3822        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3823        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
3824        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3825          forward reference subroutine calls in the group, there will be entries on
3826          the workspace list; replicate these with an appropriate increment. */
3827    
3828        else        else
3829          {          {
3830          if (repeat_min > 1)          if (repeat_min > 1)
3831            {            {
3832            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3833            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
3834              potential integer overflow. */
3835    
3836              if (lengthptr != NULL)
3837                {
3838                int delta = (repeat_min - 1)*length_prevgroup;
3839                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3840                                                                (double)INT_MAX ||
3841                    OFLOW_MAX - *lengthptr < delta)
3842                  {
3843                  *errorcodeptr = ERR20;
3844                  goto FAILED;
3845                  }
3846                *lengthptr += delta;
3847                }
3848    
3849              /* This is compiling for real */
3850    
3851              else
3852              {              {
3853              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3854              code += len;              for (i = 1; i < repeat_min; i++)
3855                  {
3856                  uschar *hc;
3857                  uschar *this_hwm = cd->hwm;
3858                  memcpy(code, previous, len);
3859                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3860                    {
3861                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3862                    cd->hwm += LINK_SIZE;
3863                    }
3864                  save_hwm = this_hwm;
3865                  code += len;
3866                  }
3867              }              }
3868            }            }
3869    
3870          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3871          }          }
3872    
# Line 2677  for (;; ptr++) Line 3874  for (;; ptr++)
3874        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3875        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,
3876        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
3877        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
3878          replicate entries on the forward reference list. */
3879    
3880        if (repeat_max >= 0)        if (repeat_max >= 0)
3881          {          {
3882          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
3883            just adjust the length as if we had. For each repetition we must add 1
3884            to the length for BRAZERO and for all but the last repetition we must
3885            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3886            paranoid checks to avoid integer overflow. */
3887    
3888            if (lengthptr != NULL && repeat_max > 0)
3889              {
3890              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3891                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3892              if ((double)repeat_max *
3893                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3894                      > (double)INT_MAX ||
3895                  OFLOW_MAX - *lengthptr < delta)
3896                {
3897                *errorcodeptr = ERR20;
3898                goto FAILED;
3899                }
3900              *lengthptr += delta;
3901              }
3902    
3903            /* This is compiling for real */
3904    
3905            else for (i = repeat_max - 1; i >= 0; i--)
3906            {            {
3907              uschar *hc;
3908              uschar *this_hwm = cd->hwm;
3909    
3910            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
3911    
3912            /* All but the final copy start a new nesting, maintaining the            /* All but the final copy start a new nesting, maintaining the
# Line 2698  for (;; ptr++) Line 3922  for (;; ptr++)
3922              }              }
3923    
3924            memcpy(code, previous, len);            memcpy(code, previous, len);
3925              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3926                {
3927                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
3928                cd->hwm += LINK_SIZE;
3929                }
3930              save_hwm = this_hwm;
3931            code += len;            code += len;
3932            }            }
3933    
# Line 2720  for (;; ptr++) Line 3950  for (;; ptr++)
3950        /* 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
3951        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
3952        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
3953        correct offset was computed above. */        correct offset was computed above.
3954    
3955        else code[-ketoffset] = OP_KETRMAX + repeat_type;        Then, when we are doing the actual compile phase, check to see whether
3956          this group is a non-atomic one that could match an empty string. If so,
3957          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
3958          that runtime checking can be done. [This check is also applied to
3959          atomic groups at runtime, but in a different way.] */
3960    
3961          else
3962            {
3963            uschar *ketcode = code - ketoffset;
3964            uschar *bracode = ketcode - GET(ketcode, 1);
3965            *ketcode = OP_KETRMAX + repeat_type;
3966            if (lengthptr == NULL && *bracode != OP_ONCE)
3967              {
3968              uschar *scode = bracode;
3969              do
3970                {
3971                if (could_be_empty_branch(scode, ketcode, utf8))
3972                  {
3973                  *bracode += OP_SBRA - OP_BRA;
3974                  break;
3975                  }
3976                scode += GET(scode, 1);
3977                }
3978              while (*scode == OP_ALT);
3979              }
3980            }
3981        }        }
3982    
3983      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2733  for (;; ptr++) Line 3988  for (;; ptr++)
3988        goto FAILED;        goto FAILED;
3989        }        }
3990    
3991      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
3992      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
3993      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
3994      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.
3995      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
3996        but the special opcodes can optimize it a bit. The repeated item starts at
3997        tempcode, not at previous, which might be the first part of a string whose
3998        (former) last char we repeated.
3999    
4000        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4001        an 'upto' may follow. We skip over an 'exact' item, and then test the
4002        length of what remains before proceeding. */
4003    
4004      if (possessive_quantifier)      if (possessive_quantifier)
4005        {        {
4006        int len = code - tempcode;        int len;
4007        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4008        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
4009        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode];
4010        tempcode[0] = OP_ONCE;        len = code - tempcode;
4011        *code++ = OP_KET;        if (len > 0) switch (*tempcode)
4012        PUTINC(code, 0, len);          {
4013        PUT(tempcode, 1, len);          case OP_STAR:  *tempcode = OP_POSSTAR; break;
4014            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
4015            case OP_QUERY: *tempcode = OP_POSQUERY; break;
4016            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4017    
4018            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4019            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4020            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4021            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4022    
4023            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4024            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4025            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4026            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4027    
4028            default:
4029            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4030            code += 1 + LINK_SIZE;
4031            len += 1 + LINK_SIZE;
4032            tempcode[0] = OP_ONCE;
4033            *code++ = OP_KET;
4034            PUTINC(code, 0, len);
4035            PUT(tempcode, 1, len);
4036            break;
4037            }
4038        }        }
4039    
4040      /* In all case we no longer have a previous item. We also set the      /* In all case we no longer have a previous item. We also set the
# Line 2761  for (;; ptr++) Line 4047  for (;; ptr++)
4047      break;      break;
4048    
4049    
4050      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
4051      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4052      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
4053      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.  */  
4054    
4055      case '(':      case '(':
4056      newoptions = options;      newoptions = options;
4057      skipbytes = 0;      skipbytes = 0;
4058        bravalue = OP_CBRA;
4059        save_hwm = cd->hwm;
4060        reset_bracount = FALSE;
4061    
4062        /* First deal with various "verbs" that can be introduced by '*'. */
4063    
4064        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4065          {
4066          int i, namelen;
4067          const char *vn = verbnames;
4068          const uschar *name = ++ptr;
4069          previous = NULL;
4070          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4071          if (*ptr == ':')
4072            {
4073            *errorcodeptr = ERR59;   /* Not supported */
4074            goto FAILED;
4075            }
4076          if (*ptr != ')')
4077            {
4078            *errorcodeptr = ERR60;
4079            goto FAILED;
4080            }
4081          namelen = ptr - name;
4082          for (i = 0; i < verbcount; i++)
4083            {
4084            if (namelen == verbs[i].len &&
4085                strncmp((char *)name, vn, namelen) == 0)
4086              {
4087              *code = verbs[i].op;
4088              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4089              break;
4090              }
4091            vn += verbs[i].len + 1;
4092            }
4093          if (i < verbcount) continue;
4094          *errorcodeptr = ERR60;
4095          goto FAILED;
4096          }
4097    
4098        /* Deal with the extended parentheses; all are introduced by '?', and the
4099        appearance of any of them means that this is not a capturing group. */
4100    
4101      if (*(++ptr) == '?')      else if (*ptr == '?')
4102        {        {
4103        int set, unset;        int i, set, unset, namelen;
4104        int *optset;        int *optset;
4105          const uschar *name;
4106          uschar *slot;
4107    
4108        switch (*(++ptr))        switch (*(++ptr))
4109          {          {
4110          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
4111          ptr++;          ptr++;
4112          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
4113            if (*ptr == 0)
4114              {
4115              *errorcodeptr = ERR18;
4116              goto FAILED;
4117              }
4118          continue;          continue;
4119    
4120          case ':':                 /* Non-extracting bracket */  
4121            /* ------------------------------------------------------------ */
4122            case '|':                 /* Reset capture count for each branch */
4123            reset_bracount = TRUE;
4124            /* Fall through */
4125    
4126            /* ------------------------------------------------------------ */
4127            case ':':                 /* Non-capturing bracket */
4128          bravalue = OP_BRA;          bravalue = OP_BRA;
4129          ptr++;          ptr++;
4130          break;          break;
4131    
4132    
4133            /* ------------------------------------------------------------ */
4134          case '(':          case '(':
4135          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4136    
4137          /* Condition to test for recursion */          /* A condition can be an assertion, a number (referring to a numbered
4138            group), a name (referring to a named group), or 'R', referring to
4139            recursion. R<digits> and R&name are also permitted for recursion tests.
4140    
4141            There are several syntaxes for testing a named group: (?(name)) is used
4142            by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
4143    
4144            There are two unfortunate ambiguities, caused by history. (a) 'R' can
4145            be the recursive thing or the name 'R' (and similarly for 'R' followed
4146            by digits), and (b) a number could be a name that consists of digits.
4147            In both cases, we look for a name first; if not found, we try the other
4148            cases. */
4149    
4150            /* For conditions that are assertions, check the syntax, and then exit
4151            the switch. This will take control down to where bracketed groups,
4152            including assertions, are processed. */
4153    
4154            if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
4155              break;
4156    
4157            /* Most other conditions use OP_CREF (a couple change to OP_RREF
4158            below), and all need to skip 3 bytes at the start of the group. */
4159    
4160          if (ptr[1] == 'R')          code[1+LINK_SIZE] = OP_CREF;
4161            skipbytes = 3;
4162            refsign = -1;
4163    
4164            /* Check for a test for recursion in a named group. */
4165    
4166            if (ptr[1] == 'R' && ptr[2] == '&')
4167            {            {
4168            code[1+LINK_SIZE] = OP_CREF;            terminator = -1;
4169            PUT2(code, 2+LINK_SIZE, CREF_RECURSE);            ptr += 2;
4170            skipbytes = 3;            code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
           ptr += 3;  
4171            }            }
4172    
4173          /* 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
4174          if a digit follows ( then there will just be digits until ) because          syntax (?(<name>) or (?('name') */
         the syntax was checked in the first pass. */  
4175    
4176          else if ((digitab[ptr[1]] && ctype_digit) != 0)          else if (ptr[1] == '<')
4177            {            {
4178            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;  
             }  
4179            ptr++;            ptr++;
           code[1+LINK_SIZE] = OP_CREF;  
           PUT2(code, 2+LINK_SIZE, condref);  
           skipbytes = 3;  
4180            }            }
4181          /* For conditions that are assertions, we just fall through, having          else if (ptr[1] == '\'')
         set bravalue above. */  
         break;  
   
         case '=':                 /* Positive lookahead */  
         bravalue = OP_ASSERT;  
         ptr++;  
         break;  
   
         case '!':                 /* Negative lookahead */  
         bravalue = OP_ASSERT_NOT;  
         ptr++;  
         break;  
   
         case '<':                 /* Lookbehinds */  
         switch (*(++ptr))  
4182            {            {
4183            case '=':               /* Positive lookbehind */            terminator = '\'';
           bravalue = OP_ASSERTBACK;  
4184            ptr++;            ptr++;
4185            break;            }
4186            else
4187              {
4188              terminator = 0;
4189              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4190              }
4191    
4192            case '!':               /* Negative lookbehind */          /* We now expect to read a name; any thing else is an error */
4193            bravalue = OP_ASSERTBACK_NOT;  
4194            ptr++;          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
4195            break;            {
4196              ptr += 1;  /* To get the right offset */
4197              *errorcodeptr = ERR28;
4198              goto FAILED;
4199            }            }
         break;  
4200    
4201          case '>':                 /* One-time brackets */          /* Read the name, but also get it as a number if it's all digits */
         bravalue = OP_ONCE;  
         ptr++;  
         break;  
4202    
4203          case 'C':                 /* Callout - may be followed by digits; */          recno = 0;
4204          previous_callout = code;  /* Save for later completion */          name = ++ptr;
4205          after_manual_callout = 1; /* Skip one item before completing */          while ((cd->ctypes[*ptr] & ctype_word) != 0)
4206          *code++ = OP_CALLOUT;     /* Already checked that the terminating */            {
4207            {                       /* closing parenthesis is present. */            if (recno >= 0)
4208            int n = 0;              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4209            while ((digitab[*(++ptr)] & ctype_digit) != 0)                recno * 10 + *ptr - '0' : -1;
4210              n = n * 10 + *ptr - '0';            ptr++;
           if (n > 255)  
             {  
             *errorcodeptr = ERR38;  
             goto FAILED;  
             }  
           *code++ = n;  
           PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */  
           PUT(code, LINK_SIZE, 0);                    /* Default length */  
           code += 2 * LINK_SIZE;  
4211            }            }
4212          previous = NULL;          namelen = ptr - name;
         continue;  
4213    
4214          case 'P':                 /* Named subpattern handling */          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
         if (*(++ptr) == '<')      /* Definition */  
4215            {            {
4216            int i, namelen;            ptr--;      /* Error offset */
4217            uschar *slot = cd->name_table;            *errorcodeptr = ERR26;
4218            const uschar *name;     /* Don't amalgamate; some compilers */            goto FAILED;
4219            name = ++ptr;           /* grumble at autoincrement in declaration */            }
4220    
4221            while (*ptr++ != '>');          /* Do no further checking in the pre-compile phase. */
           namelen = ptr - name - 1;  
4222    
4223            for (i = 0; i < cd->names_found; i++)          if (lengthptr != NULL) break;
4224    
4225            /* In the real compile we do the work of looking for the actual
4226            reference. If the string started with "+" or "-" we require the rest to
4227            be digits, in which case recno will be set. */
4228    
4229            if (refsign > 0)
4230              {
4231              if (recno <= 0)
4232              {              {
4233              int crc = memcmp(name, slot+2, namelen);              *errorcodeptr = ERR58;
4234              if (crc == 0)              goto FAILED;
4235                {              }
4236                if (slot[2+namelen] == 0)            if (refsign == '-')
4237                  {              {
4238                  *errorcodeptr = ERR43;              recno = cd->bracount - recno + 1;
4239                  goto FAILED;              if (recno <= 0)
                 }  
               crc = -1;             /* Current name is substring */  
               }  
             if (crc < 0)  
4240                {                {
4241                memmove(slot + cd->name_entry_size, slot,                *errorcodeptr = ERR15;
4242                  (cd->names_found - i) * cd->name_entry_size);                goto FAILED;
               break;  
4243                }                }
             slot += cd->name_entry_size;  
4244              }              }
4245              else recno += cd->bracount;
4246            PUT2(slot, 0, *brackets + 1);            PUT2(code, 2+LINK_SIZE, recno);
4247            memcpy(slot + 2, name, namelen);            break;
           slot[2+namelen] = 0;  
           cd->names_found++;  
           goto NUMBERED_GROUP;  
4248            }            }
4249    
4250          if (*ptr == '=' || *ptr == '>')  /* Reference or recursion */          /* Otherwise (did not start with "+" or "-"), start by looking for the
4251            name. */
4252    
4253            slot = cd->name_table;
4254            for (i = 0; i < cd->names_found; i++)
4255            {            {
4256            int i, namelen;            if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
4257            int type = *ptr++;            slot += cd->name_entry_size;
4258            const uschar *name = ptr;            }
           uschar *slot = cd->name_table;  
4259    
4260            while (*ptr != ')') ptr++;          /* Found a previous named subpattern */
4261    
4262            if (i < cd->names_found)
4263              {
4264              recno = GET2(slot, 0);
4265              PUT2(code, 2+LINK_SIZE, recno);
4266              }
4267    
4268            /* Search the pattern for a forward reference */
4269    
4270            else if ((i = find_parens(ptr, cd->bracount, name, namelen,
4271                            (options & PCRE_EXTENDED) != 0)) > 0)
4272              {
4273              PUT2(code, 2+LINK_SIZE, i);
4274              }
4275    
4276            /* If terminator == 0 it means that the name followed directly after
4277            the opening parenthesis [e.g. (?(abc)...] and in this case there are
4278            some further alternatives to try. For the cases where terminator != 0
4279            [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
4280            now checked all the possibilities, so give an error. */
4281    
4282            else if (terminator != 0)
4283              {
4284              *errorcodeptr = ERR15;
4285              goto FAILED;
4286              }
4287    
4288            /* Check for (?(R) for recursion. Allow digits after R to specify a
4289            specific group number. */
4290    
4291            else if (*name == 'R')
4292              {
4293              recno = 0;
4294              for (i = 1; i < namelen; i++)
4295                {
4296                if ((digitab[name[i]] & ctype_digit) == 0)
4297                  {
4298                  *errorcodeptr = ERR15;
4299                  goto FAILED;
4300                  }
4301                recno = recno * 10 + name[i] - '0';
4302                }
4303              if (recno == 0) recno = RREF_ANY;
4304              code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
4305              PUT2(code, 2+LINK_SIZE, recno);
4306              }
4307    
4308            /* Similarly, check for the (?(DEFINE) "condition", which is always
4309            false. */
4310    
4311            else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)
4312              {
4313              code[1+LINK_SIZE] = OP_DEF;
4314              skipbytes = 1;
4315              }
4316    
4317            /* Check for the "name" actually being a subpattern number. */
4318    
4319            else if (recno > 0)
4320              {
4321              PUT2(code, 2+LINK_SIZE, recno);
4322              }
4323    
4324            /* Either an unidentified subpattern, or a reference to (?(0) */
4325    
4326            else
4327              {
4328              *errorcodeptr = (recno == 0)? ERR35: ERR15;
4329              goto FAILED;
4330              }
4331            break;
4332    
4333    
4334            /* ------------------------------------------------------------ */
4335            case '=':                 /* Positive lookahead */
4336            bravalue = OP_ASSERT;
4337            ptr++;
4338            break;
4339    
4340    
4341            /* ------------------------------------------------------------ */
4342            case '!':                 /* Negative lookahead */
4343            ptr++;
4344            if (*ptr == ')')          /* Optimize (?!) */
4345              {
4346              *code++ = OP_FAIL;
4347              previous = NULL;
4348              continue;
4349              }
4350            bravalue = OP_ASSERT_NOT;
4351            break;
4352    
4353    
4354            /* ------------------------------------------------------------ */
4355            case '<':                 /* Lookbehind or named define */
4356            switch (ptr[1])
4357              {
4358              case '=':               /* Positive lookbehind */
4359              bravalue = OP_ASSERTBACK;
4360              ptr += 2;
4361              break;
4362    
4363              case '!':               /* Negative lookbehind */
4364              bravalue = OP_ASSERTBACK_NOT;
4365              ptr += 2;
4366              break;
4367    
4368              default:                /* Could be name define, else bad */
4369              if ((cd->ctypes[ptr[1]] & ctype_word) != 0) goto DEFINE_NAME;
4370              ptr++;                  /* Correct offset for error */
4371              *errorcodeptr = ERR24;
4372              goto FAILED;
4373              }
4374            break;
4375    
4376    
4377            /* ------------------------------------------------------------ */
4378            case '>':                 /* One-time brackets */
4379            bravalue = OP_ONCE;
4380            ptr++;
4381            break;
4382    
4383    
4384            /* ------------------------------------------------------------ */
4385            case 'C':                 /* Callout - may be followed by digits; */
4386            previous_callout = code;  /* Save for later completion */
4387            after_manual_callout = 1; /* Skip one item before completing */
4388            *code++ = OP_CALLOUT;
4389              {
4390              int n = 0;
4391              while ((digitab[*(++ptr)] & ctype_digit) != 0)
4392                n = n * 10 + *ptr - '0';
4393              if (*ptr != ')')
4394                {
4395                *errorcodeptr = ERR39;
4396                goto FAILED;
4397                }
4398              if (n > 255)
4399                {
4400                *errorcodeptr = ERR38;
4401                goto FAILED;
4402                }
4403              *code++ = n;
4404              PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */
4405              PUT(code, LINK_SIZE, 0);                    /* Default length */
4406              code += 2 * LINK_SIZE;
4407              }
4408            previous = NULL;
4409            continue;
4410    
4411    
4412            /* ------------------------------------------------------------ */
4413            case 'P':                 /* Python-style named subpattern handling */
4414            if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */
4415              {
4416              is_recurse = *ptr == '>';
4417              terminator = ')';
4418              goto NAMED_REF_OR_RECURSE;
4419              }
4420            else if (*ptr != '<')    /* Test for Python-style definition */
4421              {
4422              *errorcodeptr = ERR41;
4423              goto FAILED;
4424              }
4425            /* Fall through to handle (?P< as (?< is handled */
4426    
4427    
4428            /* ------------------------------------------------------------ */
4429            DEFINE_NAME:    /* Come here from (?< handling */
4430            case '\'':
4431              {
4432              terminator = (*ptr == '<')? '>' : '\'';