/[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 340 by ph10, Fri Apr 18 20:00:21 2008 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) 1997-2008 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 const 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 const int verbcount = sizeof(verbs)/sizeof(verbitem);
172    
173    
174    /* Tables of names of POSIX character classes and their lengths. The names are
175    now all in a single string, to reduce the number of relocations when a shared
176    library is dynamically loaded. The list of lengths is terminated by a zero
177    length entry. The first three must be alpha, lower, upper, as this is assumed
178    for handling case independence. */
179    
180    static const char posix_names[] =
181      "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"
182      "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"
183      "word\0"   "xdigit";
184    
185  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
186    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
187    
188  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
189  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
190  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
191    characters are removed, and for [:alpha:] and [:alnum:] the underscore
192    character is removed. The triples in the table consist of the base map offset,
193    second map offset or -1 if no second map, and a non-negative value for map
194    addition or a negative value for map subtraction (if there are two maps). The
195    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
196    remove vertical space characters, 2 => remove underscore. */
197    
198  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
199    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
200    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
201    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
202    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
203    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
204    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
205    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
206    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
207    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
208    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
209    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
210    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
211    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
212    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
213  };  };
214    
215    
216  /* The texts of compile-time error messages. These are "char *" because they  #define STRING(a)  # a
217  are passed to the outside world. */  #define XSTRING(s) STRING(s)
218    
219  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
220    "no error",  are passed to the outside world. Do not ever re-use any error number, because
221    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
222    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
223    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
224    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
225    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
226    simply count through to the one we want - this isn't a performance issue
227    because these strings are used only when there is a compilation error. */
228    
229    static const char error_texts[] =
230      "no error\0"
231      "\\ at end of pattern\0"
232      "\\c at end of pattern\0"
233      "unrecognized character follows \\\0"
234      "numbers out of order in {} quantifier\0"
235    /* 5 */    /* 5 */
236    "number too big in {} quantifier",    "number too big in {} quantifier\0"
237    "missing terminating ] for character class",    "missing terminating ] for character class\0"
238    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
239    "range out of order in character class",    "range out of order in character class\0"
240    "nothing to repeat",    "nothing to repeat\0"
241    /* 10 */    /* 10 */
242    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
243    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
244    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
245    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
246    "missing )",    "missing )\0"
247    /* 15 */    /* 15 */
248    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
249    "erroffset passed as NULL",    "erroffset passed as NULL\0"
250    "unknown option bit(s) set",    "unknown option bit(s) set\0"
251    "missing ) after comment",    "missing ) after comment\0"
252    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
253    /* 20 */    /* 20 */
254    "regular expression too large",    "regular expression is too large\0"
255    "failed to get memory",    "failed to get memory\0"
256    "unmatched parentheses",    "unmatched parentheses\0"
257    "internal error: code overflow",    "internal error: code overflow\0"
258    "unrecognized character after (?<",    "unrecognized character after (?<\0"
259    /* 25 */    /* 25 */
260    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
261    "malformed number after (?(",    "malformed number or name after (?(\0"
262    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
263    "assertion expected after (?(",    "assertion expected after (?(\0"
264    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
265    /* 30 */    /* 30 */
266    "unknown POSIX class name",    "unknown POSIX class name\0"
267    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
268    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
269    "spare error",    "spare error\0"  /** DEAD **/
270    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
271    /* 35 */    /* 35 */
272    "invalid condition (?(0)",    "invalid condition (?(0)\0"
273    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
274    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
275    "number after (?C is > 255",    "number after (?C is > 255\0"
276    "closing ) for (?C expected",    "closing ) for (?C expected\0"
277    /* 40 */    /* 40 */
278    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
279    "unrecognized character after (?P",    "unrecognized character after (?P\0"
280    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
281    "two named groups have the same name",    "two named subpatterns have the same name\0"
282    "invalid UTF-8 string",    "invalid UTF-8 string\0"
283    /* 45 */    /* 45 */
284    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
285    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
286    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
287  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
288      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
289      /* 50 */
290      "repeated subpattern is too long\0"    /** DEAD **/
291      "octal value is greater than \\377 (not in UTF-8 mode)\0"
292      "internal error: overran compiling workspace\0"
293      "internal error: previously-checked referenced subpattern not found\0"
294      "DEFINE group contains more than one branch\0"
295      /* 55 */
296      "repeating a DEFINE group is not allowed\0"
297      "inconsistent NEWLINE options\0"
298      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
299      "a numbered reference must not be zero\0"
300      "(*VERB) with an argument is not supported\0"
301      /* 60 */
302      "(*VERB) not recognized\0"
303      "number is too big\0"
304      "subpattern name expected\0"
305      "digit expected after (?+\0"
306      "] is an invalid data character in JavaScript compatibility mode";
307    
308    
309  /* 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 322  For convenience, we use the same bit def
322    
323  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
324    
325  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */
326  static const unsigned char digitab[] =  static const unsigned char digitab[] =
327    {    {
328    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 358  static const unsigned char digitab[] =
358    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
359    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
360    
361  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else           /* This is the "abnormal" case, for EBCDIC systems */
362  static const unsigned char digitab[] =  static const unsigned char digitab[] =
363    {    {
364    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 372  static const unsigned char digitab[] =
372    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
373    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
374    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
375    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
376    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
377    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
378    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 406  static const unsigned char ebcdic_charta
406    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
407    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
408    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
409    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
410    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
411    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
412    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 433  static const unsigned char ebcdic_charta
433  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
434    
435  static BOOL  static BOOL
436    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
437      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
438    
439    
440    
441    /*************************************************
442    *            Find an error text                  *
443    *************************************************/
444    
445    /* The error texts are now all in one long string, to save on relocations. As
446    some of the text is of unknown length, we can't use a table of offsets.
447    Instead, just count through the strings. This is not a performance issue
448    because it happens only when there has been a compilation error.
449    
450    Argument:   the error number
451    Returns:    pointer to the error string
452    */
453    
454    static const char *
455    find_error_text(int n)
456    {
457    const char *s = error_texts;
458    for (; n > 0; n--) while (*s++ != 0);
459    return s;
460    }
461    
462    
463  /*************************************************  /*************************************************
# Line 342  static BOOL Line 466  static BOOL
466    
467  /* 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
468  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
469  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
470  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
471  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,
472    ptr is pointing at the \. On exit, it is on the final character of the escape
473    sequence.
474    
475  Arguments:  Arguments:
476    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 481  Arguments:
481    
482  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
483                   negative => a special escape sequence                   negative => a special escape sequence
484                   on error, errorptr is set                   on error, errorcodeptr is set
485  */  */
486    
487  static int  static int
488  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
489    int options, BOOL isclass)    int options, BOOL isclass)
490  {  {
491  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
492    const uschar *ptr = *ptrptr + 1;
493  int c, i;  int c, i;
494    
495    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
496    ptr--;                            /* Set pointer back to the last byte */
497    
498  /* 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. */
499    
 c = *(++ptr);  
500  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
501    
502  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
503  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
504  Otherwise further processing may be required. */  Otherwise further processing may be required. */
505    
506  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
507  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */
508  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
509    
510  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
511  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
512  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
513  #endif  #endif
514    
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 517  else if ((i = escapes[c - 0x48]) != 0)
517  else  else
518    {    {
519    const uschar *oldptr;    const uschar *oldptr;
520      BOOL braced, negated;
521    
522    switch (c)    switch (c)
523      {      {
524      /* 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 532  else
532      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
533      break;      break;
534    
535        /* \g must be followed by one of a number of specific things:
536    
537        (1) A number, either plain or braced. If positive, it is an absolute
538        backreference. If negative, it is a relative backreference. This is a Perl
539        5.10 feature.
540    
541        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
542        is part of Perl's movement towards a unified syntax for back references. As
543        this is synonymous with \k{name}, we fudge it up by pretending it really
544        was \k.
545    
546        (3) For Oniguruma compatibility we also support \g followed by a name or a
547        number either in angle brackets or in single quotes. However, these are
548        (possibly recursive) subroutine calls, _not_ backreferences. Just return
549        the -ESC_g code (cf \k). */
550    
551        case 'g':
552        if (ptr[1] == '<' || ptr[1] == '\'')
553          {
554          c = -ESC_g;
555          break;
556          }
557    
558        /* Handle the Perl-compatible cases */
559    
560        if (ptr[1] == '{')
561          {
562          const uschar *p;
563          for (p = ptr+2; *p != 0 && *p != '}'; p++)
564            if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;
565          if (*p != 0 && *p != '}')
566            {
567            c = -ESC_k;
568            break;
569            }
570          braced = TRUE;
571          ptr++;
572          }
573        else braced = FALSE;
574    
575        if (ptr[1] == '-')
576          {
577          negated = TRUE;
578          ptr++;
579          }
580        else negated = FALSE;
581    
582        c = 0;
583        while ((digitab[ptr[1]] & ctype_digit) != 0)
584          c = c * 10 + *(++ptr) - '0';
585    
586        if (c < 0)   /* Integer overflow */
587          {
588          *errorcodeptr = ERR61;
589          break;
590          }
591    
592        if (braced && *(++ptr) != '}')
593          {
594          *errorcodeptr = ERR57;
595          break;
596          }
597    
598        if (c == 0)
599          {
600          *errorcodeptr = ERR58;
601          break;
602          }
603    
604        if (negated)
605          {
606          if (c > bracount)
607            {
608            *errorcodeptr = ERR15;
609            break;
610            }
611          c = bracount - (c - 1);
612          }
613    
614        c = -(ESC_REF + c);
615        break;
616    
617      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
618      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
619      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 422  else Line 635  else
635        c -= '0';        c -= '0';
636        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
637          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
638          if (c < 0)    /* Integer overflow */
639            {
640            *errorcodeptr = ERR61;
641            break;
642            }
643        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
644          {          {
645          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 442  else Line 660  else
660        }        }
661    
662      /* \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
663      larger first octal digit. */      larger first octal digit. The original code used just to take the least
664        significant 8 bits of octal numbers (I think this is what early Perls used
665        to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
666        than 3 octal digits. */
667    
668      case '0':      case '0':
669      c -= '0';      c -= '0';
670      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')
671          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - '0';
672      c &= 255;     /* Take least significant 8 bits */      if (!utf8 && c > 255) *errorcodeptr = ERR51;
673      break;      break;
674    
675      /* \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
676      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
677        treated as a data character. */
678    
679      case 'x':      case 'x':
680  #ifdef SUPPORT_UTF8      if (ptr[1] == '{')
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
681        {        {
682        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
683        register int count = 0;        int count = 0;
684    
685        c = 0;        c = 0;
686        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
687          {          {
688          int cc = *pt++;          register int cc = *pt++;
689            if (c == 0 && cc == '0') continue;     /* Leading zeroes */
690          count++;          count++;
691  #if !EBCDIC    /* ASCII coding */  
692    #ifndef EBCDIC  /* ASCII coding */
693          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= 'a') cc -= 32;               /* Convert to upper case */
694          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));
695  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
696          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */
697          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));
698  #endif  #endif
699          }          }
700    
701        if (*pt == '}')        if (*pt == '}')
702          {          {
703          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
704          ptr = pt;          ptr = pt;
705          break;          break;
706          }          }
707    
708        /* 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
709        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
710        }        }
 #endif  
711    
712      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
713    
714      c = 0;      c = 0;
715      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
716        {        {
717        int cc;                               /* Some compilers don't like ++ */        int cc;                               /* Some compilers don't like ++ */
718        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                        /* in initializers */
719  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
720        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= 'a') cc -= 32;              /* Convert to upper case */
721        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));
722  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
723        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= 'z') cc += 64;              /* Convert to upper case */
724        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));
725  #endif  #endif
726        }        }
727      break;      break;
728    
729      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
730        This coding is ASCII-specific, but then the whole concept of \cx is
731        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
732    
733      case 'c':      case 'c':
734      c = *(++ptr);      c = *(++ptr);
735      if (c == 0)      if (c == 0)
736        {        {
737        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
738        return 0;        break;
739        }        }
740    
741      /* 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 */  
742      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= 'a' && c <= 'z') c -= 32;
743      c ^= 0x40;      c ^= 0x40;
744  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
745      if (c >= 'a' && c <= 'z') c += 64;      if (c >= 'a' && c <= 'z') c += 64;
746      c ^= 0xC0;      c ^= 0xC0;
747  #endif  #endif
748      break;      break;
749    
750      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
751      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
752      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
753      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
754      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
755    
756      default:      default:
757      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 560  escape sequence. Line 783  escape sequence.
783  Argument:  Argument:
784    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
785    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
786      dptr           points to an int that is set to the detailed property value
787    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
788    
789  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
790  */  */
791    
792  static int  static int
793  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
794  {  {
795  int c, i, bot, top;  int c, i, bot, top;
796  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
797  char name[4];  char name[32];
798    
799  c = *(++ptr);  c = *(++ptr);
800  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
801    
802  *negptr = FALSE;  *negptr = FALSE;
803    
804  /* \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
805  preceded by ^ for negation. */  negation. */
806    
807  if (c == '{')  if (c == '{')
808    {    {
# Line 587  if (c == '{') Line 811  if (c == '{')
811      *negptr = TRUE;      *negptr = TRUE;
812      ptr++;      ptr++;
813      }      }
814    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
815      {      {
816      c = *(++ptr);      c = *(++ptr);
817      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
818      if (c == '}') break;      if (c == '}') break;
819      name[i] = c;      name[i] = c;
820      }      }
821    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;  
     }  
822    name[i] = 0;    name[i] = 0;
823    }    }
824    
# Line 619  top = _pcre_utt_size; Line 839  top = _pcre_utt_size;
839    
840  while (bot < top)  while (bot < top)
841    {    {
842    i = (bot + top)/2;    i = (bot + top) >> 1;
843    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
844    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
845        {
846        *dptr = _pcre_utt[i].value;
847        return _pcre_utt[i].type;
848        }
849    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
850    }    }
851    
 UNKNOWN_RETURN:  
852  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
853  *ptrptr = ptr;  *ptrptr = ptr;
854  return -1;  return -1;
# Line 698  read_repeat_counts(const uschar *p, int Line 921  read_repeat_counts(const uschar *p, int
921  int min = 0;  int min = 0;
922  int max = -1;  int max = -1;
923    
924    /* Read the minimum value and do a paranoid check: a negative value indicates
925    an integer overflow. */
926    
927  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';
928    if (min < 0 || min > 65535)
929      {
930      *errorcodeptr = ERR5;
931      return p;
932      }
933    
934    /* Read the maximum value if there is one, and again do a paranoid on its size.
935    Also, max must not be less than min. */
936    
937  if (*p == '}') max = min; else  if (*p == '}') max = min; else
938    {    {
# Line 706  if (*p == '}') max = min; else Line 940  if (*p == '}') max = min; else
940      {      {
941      max = 0;      max = 0;
942      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';
943        if (max < 0 || max > 65535)
944          {
945          *errorcodeptr = ERR5;
946          return p;
947          }
948      if (max < min)      if (max < min)
949        {        {
950        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 953  if (*p == '}') max = min; else
953      }      }
954    }    }
955    
956  /* 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
957  pointer to the terminating '}'. */  '}'. */
958    
959  if (min > 65535 || max > 65535)  *minp = min;
960    *errorcodeptr = ERR5;  *maxp = max;
961  else  return p;
962    }
963    
964    
965    
966    /*************************************************
967    *       Find forward referenced subpattern       *
968    *************************************************/
969    
970    /* This function scans along a pattern's text looking for capturing
971    subpatterns, and counting them. If it finds a named pattern that matches the
972    name it is given, it returns its number. Alternatively, if the name is NULL, it
973    returns when it reaches a given numbered subpattern. This is used for forward
974    references to subpatterns. We know that if (?P< is encountered, the name will
975    be terminated by '>' because that is checked in the first pass.
976    
977    Arguments:
978      ptr          current position in the pattern
979      count        current count of capturing parens so far encountered
980      name         name to seek, or NULL if seeking a numbered subpattern
981      lorn         name length, or subpattern number if name is NULL
982      xmode        TRUE if we are in /x mode
983    
984    Returns:       the number of the named subpattern, or -1 if not found
985    */
986    
987    static int
988    find_parens(const uschar *ptr, int count, const uschar *name, int lorn,
989      BOOL xmode)
990    {
991    const uschar *thisname;
992    
993    for (; *ptr != 0; ptr++)
994    {    {
995    *minp = min;    int term;
996    *maxp = max;  
997      /* Skip over backslashed characters and also entire \Q...\E */
998    
999      if (*ptr == '\\')
1000        {
1001        if (*(++ptr) == 0) return -1;
1002        if (*ptr == 'Q') for (;;)
1003          {
1004          while (*(++ptr) != 0 && *ptr != '\\');
1005          if (*ptr == 0) return -1;
1006          if (*(++ptr) == 'E') break;
1007          }
1008        continue;
1009        }
1010    
1011      /* Skip over character classes; this logic must be similar to the way they
1012      are handled for real. If the first character is '^', skip it. Also, if the
1013      first few characters (either before or after ^) are \Q\E or \E we skip them
1014      too. This makes for compatibility with Perl. */
1015    
1016      if (*ptr == '[')
1017        {
1018        BOOL negate_class = FALSE;
1019        for (;;)
1020          {
1021          int c = *(++ptr);
1022          if (c == '\\')
1023            {
1024            if (ptr[1] == 'E') ptr++;
1025              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
1026                else break;
1027            }
1028          else if (!negate_class && c == '^')
1029            negate_class = TRUE;
1030          else break;
1031          }
1032    
1033        /* If the next character is ']', it is a data character that must be
1034        skipped. */
1035    
1036        if (ptr[1] == ']') ptr++;
1037    
1038        while (*(++ptr) != ']')
1039          {
1040          if (*ptr == 0) return -1;
1041          if (*ptr == '\\')
1042            {
1043            if (*(++ptr) == 0) return -1;
1044            if (*ptr == 'Q') for (;;)
1045              {
1046              while (*(++ptr) != 0 && *ptr != '\\');
1047              if (*ptr == 0) return -1;
1048              if (*(++ptr) == 'E') break;
1049              }
1050            continue;
1051            }
1052          }
1053        continue;
1054        }
1055    
1056      /* Skip comments in /x mode */
1057    
1058      if (xmode && *ptr == '#')
1059        {
1060        while (*(++ptr) != 0 && *ptr != '\n');
1061        if (*ptr == 0) return -1;
1062        continue;
1063        }
1064    
1065      /* An opening parens must now be a real metacharacter */
1066    
1067      if (*ptr != '(') continue;
1068      if (ptr[1] != '?' && ptr[1] != '*')
1069        {
1070        count++;
1071        if (name == NULL && count == lorn) return count;
1072        continue;
1073        }
1074    
1075      ptr += 2;
1076      if (*ptr == 'P') ptr++;                      /* Allow optional P */
1077    
1078      /* We have to disambiguate (?<! and (?<= from (?<name> */
1079    
1080      if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&
1081           *ptr != '\'')
1082        continue;
1083    
1084      count++;
1085    
1086      if (name == NULL && count == lorn) return count;
1087      term = *ptr++;
1088      if (term == '<') term = '>';
1089      thisname = ptr;
1090      while (*ptr != term) ptr++;
1091      if (name != NULL && lorn == ptr - thisname &&
1092          strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1093        return count;
1094    }    }
1095  return p;  
1096    return -1;
1097  }  }
1098    
1099    
# Line 778  for (;;) Line 1147  for (;;)
1147    
1148      case OP_CALLOUT:      case OP_CALLOUT:
1149      case OP_CREF:      case OP_CREF:
1150      case OP_BRANUMBER:      case OP_RREF:
1151        case OP_DEF:
1152      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1153      break;      break;
1154    
# Line 823  for (;;) Line 1193  for (;;)
1193    {    {
1194    int d;    int d;
1195    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1196    switch (op)    switch (op)
1197      {      {
1198        case OP_CBRA:
1199      case OP_BRA:      case OP_BRA:
1200      case OP_ONCE:      case OP_ONCE:
1201      case OP_COND:      case OP_COND:
1202      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1203      if (d < 0) return d;      if (d < 0) return d;
1204      branchlength += d;      branchlength += d;
1205      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 865  for (;;) Line 1234  for (;;)
1234      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1235    
1236      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1237      case OP_CREF:      case OP_CREF:
1238        case OP_RREF:
1239        case OP_DEF:
1240      case OP_OPT:      case OP_OPT:
1241      case OP_CALLOUT:      case OP_CALLOUT:
1242      case OP_SOD:      case OP_SOD:
# Line 884  for (;;) Line 1254  for (;;)
1254    
1255      case OP_CHAR:      case OP_CHAR:
1256      case OP_CHARNC:      case OP_CHARNC:
1257        case OP_NOT:
1258      branchlength++;      branchlength++;
1259      cc += 2;      cc += 2;
1260  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 910  for (;;) Line 1281  for (;;)
1281    
1282      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1283      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1284        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1285      cc += 4;      cc += 4;
1286      break;      break;
1287    
# Line 917  for (;;) Line 1289  for (;;)
1289    
1290      case OP_PROP:      case OP_PROP:
1291      case OP_NOTPROP:      case OP_NOTPROP:
1292      cc++;      cc += 2;
1293      /* Fall through */      /* Fall through */
1294    
1295      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 998  Returns:      pointer to the opcode for Line 1370  Returns:      pointer to the opcode for
1370  static const uschar *  static const uschar *
1371  find_bracket(const uschar *code, BOOL utf8, int number)  find_bracket(const uschar *code, BOOL utf8, int number)
1372  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1373  for (;;)  for (;;)
1374    {    {
1375    register int c = *code;    register int c = *code;
1376    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1377    else if (c > OP_BRA)  
1378      /* XCLASS is used for classes that cannot be represented just by a bit
1379      map. This includes negated single high-valued characters. The length in
1380      the table is zero; the actual length is stored in the compiled code. */
1381    
1382      if (c == OP_XCLASS) code += GET(code, 1);
1383    
1384      /* Handle capturing bracket */
1385    
1386      else if (c == OP_CBRA)
1387      {      {
1388      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1389      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1390      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1391      }      }
1392    
1393      /* Otherwise, we can get the item's length from the table, except that for
1394      repeated character types, we have to test for \p and \P, which have an extra
1395      two bytes of parameters. */
1396    
1397    else    else
1398      {      {
1399      code += _pcre_OP_lengths[c];      switch(c)
1400          {
1401          case OP_TYPESTAR:
1402          case OP_TYPEMINSTAR:
1403          case OP_TYPEPLUS:
1404          case OP_TYPEMINPLUS:
1405          case OP_TYPEQUERY:
1406          case OP_TYPEMINQUERY:
1407          case OP_TYPEPOSSTAR:
1408          case OP_TYPEPOSPLUS:
1409          case OP_TYPEPOSQUERY:
1410          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1411          break;
1412    
1413  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1414          case OP_TYPEMINUPTO:
1415          case OP_TYPEEXACT:
1416          case OP_TYPEPOSUPTO:
1417          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1418          break;
1419          }
1420    
1421      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* Add in the fixed length from the table */
1422      by a multi-byte character. The length in the table is a minimum, so we have  
1423      to scan along to skip the extra bytes. All opcodes are less than 128, so we      code += _pcre_OP_lengths[c];
1424      can use relatively efficient code. */  
1425      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1426      a multi-byte character. The length in the table is a minimum, so we have to
1427      arrange to skip the extra bytes. */
1428    
1429    #ifdef SUPPORT_UTF8
1430      if (utf8) switch(c)      if (utf8) switch(c)
1431        {        {
1432        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1434  for (;;)
1434        case OP_EXACT:        case OP_EXACT:
1435        case OP_UPTO:        case OP_UPTO:
1436        case OP_MINUPTO:        case OP_MINUPTO:
1437          case OP_POSUPTO:
1438        case OP_STAR:        case OP_STAR:
1439        case OP_MINSTAR:        case OP_MINSTAR:
1440          case OP_POSSTAR:
1441        case OP_PLUS:        case OP_PLUS:
1442        case OP_MINPLUS:        case OP_MINPLUS:
1443          case OP_POSPLUS:
1444        case OP_QUERY:        case OP_QUERY:
1445        case OP_MINQUERY:        case OP_MINQUERY:
1446        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1447        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;  
1448        break;        break;
1449        }        }
1450  #endif  #endif
# Line 1072  Returns:      pointer to the opcode for Line 1471  Returns:      pointer to the opcode for
1471  static const uschar *  static const uschar *
1472  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1473  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1474  for (;;)  for (;;)
1475    {    {
1476    register int c = *code;    register int c = *code;
1477    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1478    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1479    else if (c > OP_BRA)  
1480      {    /* XCLASS is used for classes that cannot be represented just by a bit
1481      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1482      }    the table is zero; the actual length is stored in the compiled code. */
1483    
1484      if (c == OP_XCLASS) code += GET(code, 1);
1485    
1486      /* Otherwise, we can get the item's length from the table, except that for
1487      repeated character types, we have to test for \p and \P, which have an extra
1488      two bytes of parameters. */
1489    
1490    else    else
1491      {      {
1492      code += _pcre_OP_lengths[c];      switch(c)
1493          {
1494          case OP_TYPESTAR:
1495          case OP_TYPEMINSTAR:
1496          case OP_TYPEPLUS:
1497          case OP_TYPEMINPLUS:
1498          case OP_TYPEQUERY:
1499          case OP_TYPEMINQUERY:
1500          case OP_TYPEPOSSTAR:
1501          case OP_TYPEPOSPLUS:
1502          case OP_TYPEPOSQUERY:
1503          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1504          break;
1505    
1506  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1507          case OP_TYPEUPTO:
1508          case OP_TYPEMINUPTO:
1509          case OP_TYPEEXACT:
1510          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1511          break;
1512          }
1513    
1514        /* Add in the fixed length from the table */
1515    
1516        code += _pcre_OP_lengths[c];
1517    
1518      /* 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
1519      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
1520      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. */  
1521    
1522    #ifdef SUPPORT_UTF8
1523      if (utf8) switch(c)      if (utf8) switch(c)
1524        {        {
1525        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1527  for (;;)
1527        case OP_EXACT:        case OP_EXACT:
1528        case OP_UPTO:        case OP_UPTO:
1529        case OP_MINUPTO:        case OP_MINUPTO:
1530          case OP_POSUPTO:
1531        case OP_STAR:        case OP_STAR:
1532        case OP_MINSTAR:        case OP_MINSTAR:
1533          case OP_POSSTAR:
1534        case OP_PLUS:        case OP_PLUS:
1535        case OP_MINPLUS:        case OP_MINPLUS:
1536          case OP_POSPLUS:
1537        case OP_QUERY:        case OP_QUERY:
1538        case OP_MINQUERY:        case OP_MINQUERY:
1539        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1540        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;  
1541        break;        break;
1542        }        }
1543  #endif  #endif
# Line 1132  for (;;) Line 1552  for (;;)
1552  *************************************************/  *************************************************/
1553    
1554  /* 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
1555  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()
1556  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
1557  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
1558  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1559    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1560    bracket whose current branch will already have been scanned.
1561    
1562  Arguments:  Arguments:
1563    code        points to start of search    code        points to start of search
# Line 1149  static BOOL Line 1571  static BOOL
1571  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1572  {  {
1573  register int c;  register int c;
1574  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);
1575       code < endcode;       code < endcode;
1576       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1577    {    {
# Line 1157  for (code = first_significant_code(code Line 1579  for (code = first_significant_code(code
1579    
1580    c = *code;    c = *code;
1581    
1582    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1583      first_significant_code() with a TRUE final argument. */
1584    
1585      if (c == OP_ASSERT)
1586        {
1587        do code += GET(code, 1); while (*code == OP_ALT);
1588        c = *code;
1589        continue;
1590        }
1591    
1592      /* Groups with zero repeats can of course be empty; skip them. */
1593    
1594      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1595        {
1596        code += _pcre_OP_lengths[c];
1597        do code += GET(code, 1); while (*code == OP_ALT);
1598        c = *code;
1599        continue;
1600        }
1601    
1602      /* For other groups, scan the branches. */
1603    
1604      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1605      {      {
1606      BOOL empty_branch;      BOOL empty_branch;
1607      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 1617  for (code = first_significant_code(code
1617        }        }
1618      while (*code == OP_ALT);      while (*code == OP_ALT);
1619      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
     code += 1 + LINK_SIZE;  
1620      c = *code;      c = *code;
1621        continue;
1622      }      }
1623    
1624    else switch (c)    /* Handle the other opcodes */
1625    
1626      switch (c)
1627      {      {
1628      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1629        cannot be represented just by a bit map. This includes negated single
1630        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1631        actual length is stored in the compiled code, so we must update "code"
1632        here. */
1633    
1634  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1635      case OP_XCLASS:      case OP_XCLASS:
1636      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1637      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1638  #endif  #endif
1639    
# Line 1233  for (code = first_significant_code(code Line 1683  for (code = first_significant_code(code
1683      case OP_NOT:      case OP_NOT:
1684      case OP_PLUS:      case OP_PLUS:
1685      case OP_MINPLUS:      case OP_MINPLUS:
1686        case OP_POSPLUS:
1687      case OP_EXACT:      case OP_EXACT:
1688      case OP_NOTPLUS:      case OP_NOTPLUS:
1689      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1690        case OP_NOTPOSPLUS:
1691      case OP_NOTEXACT:      case OP_NOTEXACT:
1692      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1693      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1694        case OP_TYPEPOSPLUS:
1695      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1696      return FALSE;      return FALSE;
1697    
1698        /* These are going to continue, as they may be empty, but we have to
1699        fudge the length for the \p and \P cases. */
1700    
1701        case OP_TYPESTAR:
1702        case OP_TYPEMINSTAR:
1703        case OP_TYPEPOSSTAR:
1704        case OP_TYPEQUERY:
1705        case OP_TYPEMINQUERY:
1706        case OP_TYPEPOSQUERY:
1707        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1708        break;
1709    
1710        /* Same for these */
1711    
1712        case OP_TYPEUPTO:
1713        case OP_TYPEMINUPTO:
1714        case OP_TYPEPOSUPTO:
1715        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1716        break;
1717    
1718      /* End of branch */      /* End of branch */
1719    
1720      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 1723  for (code = first_significant_code(code
1723      case OP_ALT:      case OP_ALT:
1724      return TRUE;      return TRUE;
1725    
1726      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1727      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1728    
1729  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1730      case OP_STAR:      case OP_STAR:
1731      case OP_MINSTAR:      case OP_MINSTAR:
1732        case OP_POSSTAR:
1733      case OP_QUERY:      case OP_QUERY:
1734      case OP_MINQUERY:      case OP_MINQUERY:
1735        case OP_POSQUERY:
1736      case OP_UPTO:      case OP_UPTO:
1737      case OP_MINUPTO:      case OP_MINUPTO:
1738        case OP_POSUPTO:
1739      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1740      break;      break;
1741  #endif  #endif
# Line 1308  return TRUE; Line 1784  return TRUE;
1784  *************************************************/  *************************************************/
1785    
1786  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1787  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
1788  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1789  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1790    
1791    Originally, this function only recognized a sequence of letters between the
1792    terminators, but it seems that Perl recognizes any sequence of characters,
1793    though of course unknown POSIX names are subsequently rejected. Perl gives an
1794    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1795    didn't consider this to be a POSIX class. Likewise for [:1234:].
1796    
1797    The problem in trying to be exactly like Perl is in the handling of escapes. We
1798    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1799    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1800    below handles the special case of \], but does not try to do any other escape
1801    processing. This makes it different from Perl for cases such as [:l\ower:]
1802    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1803    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1804    I think.
1805    
1806  Argument:  Arguments:
1807    ptr      pointer to the initial [    ptr      pointer to the initial [
1808    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1809    
1810  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1811  */  */
1812    
1813  static BOOL  static BOOL
1814  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1815  {  {
1816  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1817  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1818  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1819    {    {
1820    *endptr = ptr;    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1821    return TRUE;      {
1822        if (*ptr == ']') return FALSE;
1823        if (*ptr == terminator && ptr[1] == ']')
1824          {
1825          *endptr = ptr;
1826          return TRUE;
1827          }
1828        }
1829    }    }
1830  return FALSE;  return FALSE;
1831  }  }
# Line 1355  Returns:     a value representing the na Line 1850  Returns:     a value representing the na
1850  static int  static int
1851  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1852  {  {
1853    const char *pn = posix_names;
1854  register int yield = 0;  register int yield = 0;
1855  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1856    {    {
1857    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1858      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1859      pn += posix_name_lengths[yield] + 1;
1860    yield++;    yield++;
1861    }    }
1862  return -1;  return -1;
# Line 1374  return -1; Line 1871  return -1;
1871  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
1872  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
1873  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
1874  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
1875  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
1876  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
1877  offsets adjusted. That is the job of this function. Before it is called, the  have their offsets adjusted. That one of the jobs of this function. Before it
1878  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
1879    OP_END.
1880    
1881    This function has been extended with the possibility of forward references for
1882    recursions and subroutine calls. It must also check the list of such references
1883    for the group we are dealing with. If it finds that one of the recursions in
1884    the current group is on this list, it adjusts the offset in the list, not the
1885    value in the reference (which is a group number).
1886    
1887  Arguments:  Arguments:
1888    group      points to the start of the group    group      points to the start of the group
1889    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1890    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1891    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1892      save_hwm   the hwm forward reference pointer at the start of the group
1893    
1894  Returns:     nothing  Returns:     nothing
1895  */  */
1896    
1897  static void  static void
1898  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1899      uschar *save_hwm)
1900  {  {
1901  uschar *ptr = group;  uschar *ptr = group;
1902    
1903  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1904    {    {
1905    int offset = GET(ptr, 1);    int offset;
1906    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1907    
1908      /* See if this recursion is on the forward reference list. If so, adjust the
1909      reference. */
1910    
1911      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1912        {
1913        offset = GET(hc, 0);
1914        if (cd->start_code + offset == ptr + 1)
1915          {
1916          PUT(hc, 0, offset + adjust);
1917          break;
1918          }
1919        }
1920    
1921      /* Otherwise, adjust the recursion offset if it's after the start of this
1922      group. */
1923    
1924      if (hc >= cd->hwm)
1925        {
1926        offset = GET(ptr, 1);
1927        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1928        }
1929    
1930    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1931    }    }
1932  }  }
# Line 1475  Yield:        TRUE when range returned; Line 2005  Yield:        TRUE when range returned;
2005  */  */
2006    
2007  static BOOL  static BOOL
2008  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2009      unsigned int *odptr)
2010  {  {
2011  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2012    
2013  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2014    {    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2015    
2016  if (c > d) return FALSE;  if (c > d) return FALSE;
2017    
# Line 1492  next = othercase + 1; Line 2020  next = othercase + 1;
2020    
2021  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2022    {    {
2023    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (_pcre_ucp_othercase(c) != next) break;
         othercase != next)  
     break;  
2024    next++;    next++;
2025    }    }
2026    
# Line 1506  return TRUE; Line 2032  return TRUE;
2032  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2033    
2034    
2035    
2036  /*************************************************  /*************************************************
2037  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
2038  *************************************************/  *************************************************/
2039    
2040  /* 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
2041  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
2042  bits.  sense to automatically possessify the repeated item.
2043    
2044  Arguments:  Arguments:
2045    optionsptr     pointer to the option bits    op_code       the repeated op code
2046    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
2047    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
2048    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
2049    errorcodeptr   points to error code variable    ptr           next character in pattern
2050    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
2051    reqbyteptr     set to the last literal character required, else < 0    cd            contains pointers to tables etc.
2052    bcptr          points to current branch chain  
2053    cd             contains pointers to tables etc.  Returns:        TRUE if possessifying is wanted
2054    */
2055    
2056    static BOOL
2057    check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2058      const uschar *ptr, int options, compile_data *cd)
2059    {
2060    int next;
2061    
2062    /* Skip whitespace and comments in extended mode */
2063    
2064    if ((options & PCRE_EXTENDED) != 0)
2065      {
2066      for (;;)
2067        {
2068        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2069        if (*ptr == '#')
2070          {
2071          while (*(++ptr) != 0)
2072            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2073          }
2074        else break;
2075        }
2076      }
2077    
2078    /* If the next item is one that we can handle, get its value. A non-negative
2079    value is a character, a negative value is an escape value. */
2080    
2081    if (*ptr == '\\')
2082      {
2083      int temperrorcode = 0;
2084      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2085      if (temperrorcode != 0) return FALSE;
2086      ptr++;    /* Point after the escape sequence */
2087      }
2088    
2089    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2090      {
2091    #ifdef SUPPORT_UTF8
2092      if (utf8) { GETCHARINC(next, ptr); } else
2093    #endif
2094      next = *ptr++;
2095      }
2096    
2097    else return FALSE;
2098    
2099    /* Skip whitespace and comments in extended mode */
2100    
2101    if ((options & PCRE_EXTENDED) != 0)
2102      {
2103      for (;;)
2104        {
2105        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2106        if (*ptr == '#')
2107          {
2108          while (*(++ptr) != 0)
2109            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2110          }
2111        else break;
2112        }
2113      }
2114    
2115    /* If the next thing is itself optional, we have to give up. */
2116    
2117    if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
2118      return FALSE;
2119    
2120    /* Now compare the next item with the previous opcode. If the previous is a
2121    positive single character match, "item" either contains the character or, if
2122    "item" is greater than 127 in utf8 mode, the character's bytes are in
2123    utf8_char. */
2124    
2125    
2126    /* Handle cases when the next item is a character. */
2127    
2128    if (next >= 0) switch(op_code)
2129      {
2130      case OP_CHAR:
2131    #ifdef SUPPORT_UTF8
2132      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2133    #endif
2134      return item != next;
2135    
2136      /* For CHARNC (caseless character) we must check the other case. If we have
2137      Unicode property support, we can use it to test the other case of
2138      high-valued characters. */
2139    
2140      case OP_CHARNC:
2141    #ifdef SUPPORT_UTF8
2142      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2143    #endif
2144      if (item == next) return FALSE;
2145    #ifdef SUPPORT_UTF8
2146      if (utf8)
2147        {
2148        unsigned int othercase;
2149        if (next < 128) othercase = cd->fcc[next]; else
2150    #ifdef SUPPORT_UCP
2151        othercase = _pcre_ucp_othercase((unsigned int)next);
2152    #else
2153        othercase = NOTACHAR;
2154    #endif
2155        return (unsigned int)item != othercase;
2156        }
2157      else
2158    #endif  /* SUPPORT_UTF8 */
2159      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2160    
2161      /* For OP_NOT, "item" must be a single-byte character. */
2162    
2163      case OP_NOT:
2164      if (item == next) return TRUE;
2165      if ((options & PCRE_CASELESS) == 0) return FALSE;
2166    #ifdef SUPPORT_UTF8
2167      if (utf8)
2168        {
2169        unsigned int othercase;
2170        if (next < 128) othercase = cd->fcc[next]; else
2171    #ifdef SUPPORT_UCP
2172        othercase = _pcre_ucp_othercase(next);
2173    #else
2174        othercase = NOTACHAR;
2175    #endif
2176        return (unsigned int)item == othercase;
2177        }
2178      else
2179    #endif  /* SUPPORT_UTF8 */
2180      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2181    
2182      case OP_DIGIT:
2183      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2184    
2185      case OP_NOT_DIGIT:
2186      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2187    
2188      case OP_WHITESPACE:
2189      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2190    
2191      case OP_NOT_WHITESPACE:
2192      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2193    
2194      case OP_WORDCHAR:
2195      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2196    
2197      case OP_NOT_WORDCHAR:
2198      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2199    
2200      case OP_HSPACE:
2201      case OP_NOT_HSPACE:
2202      switch(next)
2203        {
2204        case 0x09:
2205        case 0x20:
2206        case 0xa0:
2207        case 0x1680:
2208        case 0x180e:
2209        case 0x2000:
2210        case 0x2001:
2211        case 0x2002:
2212        case 0x2003:
2213        case 0x2004:
2214        case 0x2005:
2215        case 0x2006:
2216        case 0x2007:
2217        case 0x2008:
2218        case 0x2009:
2219        case 0x200A:
2220        case 0x202f:
2221        case 0x205f:
2222        case 0x3000:
2223        return op_code != OP_HSPACE;
2224        default:
2225        return op_code == OP_HSPACE;
2226        }
2227    
2228      case OP_VSPACE:
2229      case OP_NOT_VSPACE:
2230      switch(next)
2231        {
2232        case 0x0a:
2233        case 0x0b:
2234        case 0x0c:
2235        case 0x0d:
2236        case 0x85:
2237        case 0x2028:
2238        case 0x2029:
2239        return op_code != OP_VSPACE;
2240        default:
2241        return op_code == OP_VSPACE;
2242        }
2243    
2244      default:
2245      return FALSE;
2246      }
2247    
2248    
2249    /* Handle the case when the next item is \d, \s, etc. */
2250    
2251    switch(op_code)
2252      {
2253      case OP_CHAR:
2254      case OP_CHARNC:
2255    #ifdef SUPPORT_UTF8
2256      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2257    #endif
2258      switch(-next)
2259        {
2260        case ESC_d:
2261        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2262    
2263        case ESC_D:
2264        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2265    
2266        case ESC_s:
2267        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2268    
2269        case ESC_S:
2270        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2271    
2272        case ESC_w:
2273        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2274    
2275        case ESC_W:
2276        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2277    
2278        case ESC_h:
2279        case ESC_H:
2280        switch(item)
2281          {
2282          case 0x09:
2283          case 0x20:
2284          case 0xa0:
2285          case 0x1680:
2286          case 0x180e:
2287          case 0x2000:
2288          case 0x2001:
2289          case 0x2002:
2290          case 0x2003:
2291          case 0x2004:
2292          case 0x2005:
2293          case 0x2006:
2294          case 0x2007:
2295          case 0x2008:
2296          case 0x2009:
2297          case 0x200A:
2298          case 0x202f:
2299          case 0x205f:
2300          case 0x3000:
2301          return -next != ESC_h;
2302          default:
2303          return -next == ESC_h;
2304          }
2305    
2306        case ESC_v:
2307        case ESC_V:
2308        switch(item)
2309          {
2310          case 0x0a:
2311          case 0x0b:
2312          case 0x0c:
2313          case 0x0d:
2314          case 0x85:
2315          case 0x2028:
2316          case 0x2029:
2317          return -next != ESC_v;
2318          default:
2319          return -next == ESC_v;
2320          }
2321    
2322        default:
2323        return FALSE;
2324        }
2325    
2326      case OP_DIGIT:
2327      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2328             next == -ESC_h || next == -ESC_v;
2329    
2330      case OP_NOT_DIGIT:
2331      return next == -ESC_d;
2332    
2333      case OP_WHITESPACE:
2334      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2335    
2336      case OP_NOT_WHITESPACE:
2337      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2338    
2339      case OP_HSPACE:
2340      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2341    
2342      case OP_NOT_HSPACE:
2343      return next == -ESC_h;
2344    
2345      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2346      case OP_VSPACE:
2347      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2348    
2349      case OP_NOT_VSPACE:
2350      return next == -ESC_v;
2351    
2352      case OP_WORDCHAR:
2353      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2354    
2355      case OP_NOT_WORDCHAR:
2356      return next == -ESC_w || next == -ESC_d;
2357    
2358      default:
2359      return FALSE;
2360      }
2361    
2362    /* Control does not reach here */
2363    }
2364    
2365    
2366    
2367    /*************************************************
2368    *           Compile one branch                   *
2369    *************************************************/
2370    
2371    /* Scan the pattern, compiling it into the a vector. If the options are
2372    changed during the branch, the pointer is used to change the external options
2373    bits. This function is used during the pre-compile phase when we are trying
2374    to find out the amount of memory needed, as well as during the real compile
2375    phase. The value of lengthptr distinguishes the two phases.
2376    
2377    Arguments:
2378      optionsptr     pointer to the option bits
2379      codeptr        points to the pointer to the current code point
2380      ptrptr         points to the current pattern pointer
2381      errorcodeptr   points to error code variable
2382      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2383      reqbyteptr     set to the last literal character required, else < 0
2384      bcptr          points to current branch chain
2385      cd             contains pointers to tables etc.
2386      lengthptr      NULL during the real compile phase
2387                     points to length accumulator during pre-compile phase
2388    
2389  Returns:         TRUE on success  Returns:         TRUE on success
2390                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2391  */  */
2392    
2393  static BOOL  static BOOL
2394  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2395    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2396    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2397  {  {
2398  int repeat_type, op_type;  int repeat_type, op_type;
2399  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 1541  int greedy_default, greedy_non_default; Line 2402  int greedy_default, greedy_non_default;
2402  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2403  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2404  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2405  int options = *optionsptr;  int options = *optionsptr;
2406  int after_manual_callout = 0;  int after_manual_callout = 0;
2407    int length_prevgroup = 0;
2408  register int c;  register int c;
2409  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2410    uschar *last_code = code;
2411    uschar *orig_code = code;
2412  uschar *tempcode;  uschar *tempcode;
2413  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2414  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
# Line 1553  const uschar *ptr = *ptrptr; Line 2416  const uschar *ptr = *ptrptr;
2416  const uschar *tempptr;  const uschar *tempptr;
2417  uschar *previous = NULL;  uschar *previous = NULL;
2418  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2419    uschar *save_hwm = NULL;
2420  uschar classbits[32];  uschar classbits[32];
2421    
2422  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2423  BOOL class_utf8;  BOOL class_utf8;
2424  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2425  uschar *class_utf8data;  uschar *class_utf8data;
2426    uschar *class_utf8data_base;
2427  uschar utf8_char[6];  uschar utf8_char[6];
2428  #else  #else
2429  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2430    uschar *utf8_char = NULL;
2431    #endif
2432    
2433    #ifdef DEBUG
2434    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2435  #endif  #endif
2436    
2437  /* Set up the default and non-default settings for greediness */  /* Set up the default and non-default settings for greediness */
# Line 1593  req_caseopt = ((options & PCRE_CASELESS) Line 2463  req_caseopt = ((options & PCRE_CASELESS)
2463  for (;; ptr++)  for (;; ptr++)
2464    {    {
2465    BOOL negate_class;    BOOL negate_class;
2466      BOOL should_flip_negation;
2467    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2468    BOOL is_quantifier;    BOOL is_quantifier;
2469      BOOL is_recurse;
2470      BOOL reset_bracount;
2471    int class_charcount;    int class_charcount;
2472    int class_lastchar;    int class_lastchar;
2473    int newoptions;    int newoptions;
2474    int recno;    int recno;
2475      int refsign;
2476    int skipbytes;    int skipbytes;
2477    int subreqbyte;    int subreqbyte;
2478    int subfirstbyte;    int subfirstbyte;
2479      int terminator;
2480    int mclength;    int mclength;
2481    uschar mcbuffer[8];    uschar mcbuffer[8];
2482    
2483    /* Next byte in the pattern */    /* Get next byte in the pattern */
2484    
2485    c = *ptr;    c = *ptr;
2486    
2487      /* If we are in the pre-compile phase, accumulate the length used for the
2488      previous cycle of this loop. */
2489    
2490      if (lengthptr != NULL)
2491        {
2492    #ifdef DEBUG
2493        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2494    #endif
2495        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2496          {
2497          *errorcodeptr = ERR52;
2498          goto FAILED;
2499          }
2500    
2501        /* There is at least one situation where code goes backwards: this is the
2502        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2503        the class is simply eliminated. However, it is created first, so we have to
2504        allow memory for it. Therefore, don't ever reduce the length at this point.
2505        */
2506    
2507        if (code < last_code) code = last_code;
2508    
2509        /* Paranoid check for integer overflow */
2510    
2511        if (OFLOW_MAX - *lengthptr < code - last_code)
2512          {
2513          *errorcodeptr = ERR20;
2514          goto FAILED;
2515          }
2516    
2517        *lengthptr += code - last_code;
2518        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2519    
2520        /* If "previous" is set and it is not at the start of the work space, move
2521        it back to there, in order to avoid filling up the work space. Otherwise,
2522        if "previous" is NULL, reset the current code pointer to the start. */
2523    
2524        if (previous != NULL)
2525          {
2526          if (previous > orig_code)
2527            {
2528            memmove(orig_code, previous, code - previous);
2529            code -= previous - orig_code;
2530            previous = orig_code;
2531            }
2532          }
2533        else code = orig_code;
2534    
2535        /* Remember where this code item starts so we can pick up the length
2536        next time round. */
2537    
2538        last_code = code;
2539        }
2540    
2541      /* In the real compile phase, just check the workspace used by the forward
2542      reference list. */
2543    
2544      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2545        {
2546        *errorcodeptr = ERR52;
2547        goto FAILED;
2548        }
2549    
2550    /* 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 */
2551    
# Line 1623  for (;; ptr++) Line 2561  for (;; ptr++)
2561        {        {
2562        if (previous_callout != NULL)        if (previous_callout != NULL)
2563          {          {
2564          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2565              complete_callout(previous_callout, ptr, cd);
2566          previous_callout = NULL;          previous_callout = NULL;
2567          }          }
2568        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1644  for (;; ptr++) Line 2583  for (;; ptr++)
2583    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2584         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2585      {      {
2586      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2587          complete_callout(previous_callout, ptr, cd);
2588      previous_callout = NULL;      previous_callout = NULL;
2589      }      }
2590    
# Line 1655  for (;; ptr++) Line 2595  for (;; ptr++)
2595      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2596      if (c == '#')      if (c == '#')
2597        {        {
2598        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2599        on the Macintosh. */          {
2600        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2601        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2602          if (*ptr != 0) continue;
2603    
2604          /* Else fall through to handle end of string */
2605          c = 0;
2606        }        }
2607      }      }
2608    
# Line 1672  for (;; ptr++) Line 2616  for (;; ptr++)
2616    
2617    switch(c)    switch(c)
2618      {      {
2619      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2620        case 0:                        /* The branch terminates at string end */
2621      case 0:      case '|':                      /* or | or ) */
     case '|':  
2622      case ')':      case ')':
2623      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2624      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2625      *codeptr = code;      *codeptr = code;
2626      *ptrptr = ptr;      *ptrptr = ptr;
2627        if (lengthptr != NULL)
2628          {
2629          if (OFLOW_MAX - *lengthptr < code - last_code)
2630            {
2631            *errorcodeptr = ERR20;
2632            goto FAILED;
2633            }
2634          *lengthptr += code - last_code;   /* To include callout length */
2635          DPRINTF((">> end branch\n"));
2636          }
2637      return TRUE;      return TRUE;
2638    
2639    
2640        /* ===================================================================*/
2641      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2642      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2643    
# Line 1711  for (;; ptr++) Line 2666  for (;; ptr++)
2666      *code++ = OP_ANY;      *code++ = OP_ANY;
2667      break;      break;
2668    
2669      /* Character classes. If the included characters are all < 255 in value, we  
2670      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
2671      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
2672      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
2673      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
2674        map as usual, then invert it at the end. However, we use a different opcode
2675        so that data characters > 255 can be handled correctly.
2676    
2677      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
2678      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,
2679      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2680      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
2681      */  
2682        In JavaScript compatibility mode, an isolated ']' causes an error. In
2683        default (Perl) mode, it is treated as a data character. */
2684    
2685        case ']':
2686        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2687          {
2688          *errorcodeptr = ERR64;
2689          goto FAILED;
2690          }
2691        goto NORMAL_CHAR;
2692    
2693      case '[':      case '[':
2694      previous = code;      previous = code;
# Line 1730  for (;; ptr++) Line 2697  for (;; ptr++)
2697      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
2698    
2699      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2700          check_posix_syntax(ptr, &tempptr, cd))          check_posix_syntax(ptr, &tempptr))
2701        {        {
2702        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;
2703        goto FAILED;        goto FAILED;
2704        }        }
2705    
2706      /* 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,
2707        if the first few characters (either before or after ^) are \Q\E or \E we
2708        skip them too. This makes for compatibility with Perl. */
2709    
2710      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2711        for (;;)
2712        {        {
       negate_class = TRUE;  
2713        c = *(++ptr);        c = *(++ptr);
2714          if (c == '\\')
2715            {
2716            if (ptr[1] == 'E') ptr++;
2717              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2718                else break;
2719            }
2720          else if (!negate_class && c == '^')
2721            negate_class = TRUE;
2722          else break;
2723        }        }
2724      else  
2725        {      /* If a class contains a negative special such as \S, we need to flip the
2726        negate_class = FALSE;      negation flag at the end, so that support for characters > 255 works
2727        }      correctly (they are all included in the class). */
2728    
2729        should_flip_negation = FALSE;
2730    
2731      /* 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
2732      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
2733      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2734    
2735      class_charcount = 0;      class_charcount = 0;
2736      class_lastchar = -1;      class_lastchar = -1;
2737    
2738        /* Initialize the 32-char bit map to all zeros. We build the map in a
2739        temporary bit of memory, in case the class contains only 1 character (less
2740        than 256), because in that case the compiled code doesn't use the bit map.
2741        */
2742    
2743        memset(classbits, 0, 32 * sizeof(uschar));
2744    
2745  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2746      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2747      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2748        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2749  #endif  #endif
2750    
     /* 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));  
   
2751      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2752      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
2753      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. */  
2754    
2755      do      if (c != 0) do
2756        {        {
2757          const uschar *oldptr;
2758    
2759  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2760        if (utf8 && c > 127)        if (utf8 && c > 127)
2761          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2762          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2763          }          }
2764    
2765          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2766          data and reset the pointer. This is so that very large classes that
2767          contain a zillion UTF-8 characters no longer overwrite the work space
2768          (which is on the stack). */
2769    
2770          if (lengthptr != NULL)
2771            {
2772            *lengthptr += class_utf8data - class_utf8data_base;
2773            class_utf8data = class_utf8data_base;
2774            }
2775    
2776  #endif  #endif
2777    
2778        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
2779    
2780        if (inescq)        if (inescq)
2781          {          {
2782          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2783            {            {
2784            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2785            ptr++;            ptr++;                            /* Skip the 'E' */
2786            continue;            continue;                         /* Carry on with next */
2787            }            }
2788          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2789          }          }
2790    
2791        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1803  for (;; ptr++) Line 2796  for (;; ptr++)
2796    
2797        if (c == '[' &&        if (c == '[' &&
2798            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2799            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr))
2800          {          {
2801          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2802          int posix_class, i;          int posix_class, taboffset, tabopt;
2803          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2804            uschar pbits[32];
2805    
2806          if (ptr[1] != ':')          if (ptr[1] != ':')
2807            {            {
# Line 1819  for (;; ptr++) Line 2813  for (;; ptr++)
2813          if (*ptr == '^')          if (*ptr == '^')
2814            {            {
2815            local_negate = TRUE;            local_negate = TRUE;
2816              should_flip_negation = TRUE;  /* Note negative special */
2817            ptr++;            ptr++;
2818            }            }
2819    
# Line 1836  for (;; ptr++) Line 2831  for (;; ptr++)
2831          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
2832            posix_class = 0;            posix_class = 0;
2833    
2834          /* 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
2835          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
2836          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
2837          white space chars afterwards. */          result into the bit map that is being built. */
2838    
2839          posix_class *= 3;          posix_class *= 3;
2840          for (i = 0; i < 3; i++)  
2841            /* Copy in the first table (always present) */
2842    
2843            memcpy(pbits, cbits + posix_class_maps[posix_class],
2844              32 * sizeof(uschar));
2845    
2846            /* If there is a second table, add or remove it as required. */
2847    
2848            taboffset = posix_class_maps[posix_class + 1];
2849            tabopt = posix_class_maps[posix_class + 2];
2850    
2851            if (taboffset >= 0)
2852            {            {
2853            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
2854            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;  
             }  
2855            else            else
2856              {              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;  
             }  
2857            }            }
2858    
2859            /* Not see if we need to remove any special characters. An option
2860            value of 1 removes vertical space and 2 removes underscore. */
2861    
2862            if (tabopt < 0) tabopt = -tabopt;
2863            if (tabopt == 1) pbits[1] &= ~0x3c;
2864              else if (tabopt == 2) pbits[11] &= 0x7f;
2865    
2866            /* Add the POSIX table or its complement into the main table that is
2867            being built and we are done. */
2868    
2869            if (local_negate)
2870              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
2871            else
2872              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
2873    
2874          ptr = tempptr + 1;          ptr = tempptr + 1;
2875          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
2876          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
2877          }          }
2878    
2879        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2880        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
2881        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.
2882        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2883        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  
2884        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2885    
2886        if (c == '\\')        if (c == '\\')
2887          {          {
2888          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2889            if (*errorcodeptr != 0) goto FAILED;
2890    
2891          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2892          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 */
2893            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2894          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2895            {            {
2896            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1890  for (;; ptr++) Line 2900  for (;; ptr++)
2900            else inescq = TRUE;            else inescq = TRUE;
2901            continue;            continue;
2902            }            }
2903            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2904    
2905          if (c < 0)          if (c < 0)
2906            {            {
2907            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2908            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2909            switch (-c)  
2910              /* Save time by not doing this in the pre-compile phase. */
2911    
2912              if (lengthptr == NULL) switch (-c)
2913              {              {
2914              case ESC_d:              case ESC_d:
2915              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
2916              continue;              continue;
2917    
2918              case ESC_D:              case ESC_D:
2919                should_flip_negation = TRUE;
2920              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2921              continue;              continue;
2922    
# Line 1910  for (;; ptr++) Line 2925  for (;; ptr++)
2925              continue;              continue;
2926    
2927              case ESC_W:              case ESC_W:
2928                should_flip_negation = TRUE;
2929              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2930              continue;              continue;
2931    
# Line 1919  for (;; ptr++) Line 2935  for (;; ptr++)
2935              continue;              continue;
2936    
2937              case ESC_S:              case ESC_S:
2938                should_flip_negation = TRUE;
2939              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2940              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2941              continue;              continue;
2942    
2943  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
2944              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
2945              case ESC_P:              }
2946    
2947              /* In the pre-compile phase, just do the recognition. */
2948    
2949              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2950                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
2951    
2952              /* We need to deal with \H, \h, \V, and \v in both phases because
2953              they use extra memory. */
2954    
2955              if (-c == ESC_h)
2956                {
2957                SETBIT(classbits, 0x09); /* VT */
2958                SETBIT(classbits, 0x20); /* SPACE */
2959                SETBIT(classbits, 0xa0); /* NSBP */
2960    #ifdef SUPPORT_UTF8
2961                if (utf8)
2962                {                {
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
2963                class_utf8 = TRUE;                class_utf8 = TRUE;
2964                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_SINGLE;
2965                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2966                *class_utf8data++ = property;                *class_utf8data++ = XCL_SINGLE;
2967                class_charcount -= 2;   /* Not a < 256 character */                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2968                  *class_utf8data++ = XCL_RANGE;
2969                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2970                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2971                  *class_utf8data++ = XCL_SINGLE;
2972                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2973                  *class_utf8data++ = XCL_SINGLE;
2974                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2975                  *class_utf8data++ = XCL_SINGLE;
2976                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2977                }                }
             continue;  
2978  #endif  #endif
2979                continue;
2980                }
2981    
2982              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_H)
2983              strict mode. By default, for compatibility with Perl, they are              {
2984              treated as literals. */              for (c = 0; c < 32; c++)
2985                  {
2986                  int x = 0xff;
2987                  switch (c)
2988                    {
2989                    case 0x09/8: x ^= 1 << (0x09%8); break;
2990                    case 0x20/8: x ^= 1 << (0x20%8); break;
2991                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2992                    default: break;
2993                    }
2994                  classbits[c] |= x;
2995                  }
2996    
2997              default:  #ifdef SUPPORT_UTF8
2998              if ((options & PCRE_EXTRA) != 0)              if (utf8)
2999                {                {
3000                *errorcodeptr = ERR7;                class_utf8 = TRUE;
3001                goto FAILED;                *class_utf8data++ = XCL_RANGE;
3002                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3003                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3004                  *class_utf8data++ = XCL_RANGE;
3005                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3006                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3007                  *class_utf8data++ = XCL_RANGE;
3008                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3009                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3010                  *class_utf8data++ = XCL_RANGE;
3011                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3012                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3013                  *class_utf8data++ = XCL_RANGE;
3014                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3015                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3016                  *class_utf8data++ = XCL_RANGE;
3017                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3018                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3019                  *class_utf8data++ = XCL_RANGE;
3020                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3021                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3022                }                }
3023              c = *ptr;              /* The final character */  #endif
3024              class_charcount -= 2;  /* Undo the default count from above */              continue;
3025              }              }
           }  
3026    
3027          /* Fall through if we have a single character (c >= 0). This may be            if (-c == ESC_v)
3028          > 256 in UTF-8 mode. */              {
3029                SETBIT(classbits, 0x0a); /* LF */
3030                SETBIT(classbits, 0x0b); /* VT */
3031                SETBIT(classbits, 0x0c); /* FF */
3032                SETBIT(classbits, 0x0d); /* CR */
3033                SETBIT(classbits, 0x85); /* NEL */
3034    #ifdef SUPPORT_UTF8
3035                if (utf8)
3036                  {
3037                  class_utf8 = TRUE;
3038                  *class_utf8data++ = XCL_RANGE;
3039                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3040                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3041                  }
3042    #endif
3043                continue;
3044                }
3045    
3046          }   /* End of backslash handling */            if (-c == ESC_V)
3047                {
3048                for (c = 0; c < 32; c++)
3049                  {
3050                  int x = 0xff;
3051                  switch (c)
3052                    {
3053                    case 0x0a/8: x ^= 1 << (0x0a%8);
3054                                 x ^= 1 << (0x0b%8);
3055                                 x ^= 1 << (0x0c%8);
3056                                 x ^= 1 << (0x0d%8);
3057                                 break;
3058                    case 0x85/8: x ^= 1 << (0x85%8); break;
3059                    default: break;
3060                    }
3061                  classbits[c] |= x;
3062                  }
3063    
3064        /* A single character may be followed by '-' to form a range. However,  #ifdef SUPPORT_UTF8
3065                if (utf8)
3066                  {
3067                  class_utf8 = TRUE;
3068                  *class_utf8data++ = XCL_RANGE;
3069                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3070                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3071                  *class_utf8data++ = XCL_RANGE;
3072                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3073                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3074                  }
3075    #endif
3076                continue;
3077                }
3078    
3079              /* We need to deal with \P and \p in both phases. */
3080    
3081    #ifdef SUPPORT_UCP
3082              if (-c == ESC_p || -c == ESC_P)
3083                {
3084                BOOL negated;
3085                int pdata;
3086                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3087                if (ptype < 0) goto FAILED;
3088                class_utf8 = TRUE;
3089                *class_utf8data++ = ((-c == ESC_p) != negated)?
3090                  XCL_PROP : XCL_NOTPROP;
3091                *class_utf8data++ = ptype;
3092                *class_utf8data++ = pdata;
3093                class_charcount -= 2;   /* Not a < 256 character */
3094                continue;
3095                }
3096    #endif
3097              /* Unrecognized escapes are faulted if PCRE is running in its
3098              strict mode. By default, for compatibility with Perl, they are
3099              treated as literals. */
3100    
3101              if ((options & PCRE_EXTRA) != 0)
3102                {
3103                *errorcodeptr = ERR7;
3104                goto FAILED;
3105                }
3106    
3107              class_charcount -= 2;  /* Undo the default count from above */
3108              c = *ptr;              /* Get the final character and fall through */
3109              }
3110    
3111            /* Fall through if we have a single character (c >= 0). This may be
3112            greater than 256 in UTF-8 mode. */
3113    
3114            }   /* End of backslash handling */
3115    
3116          /* A single character may be followed by '-' to form a range. However,
3117        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
3118        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3119          entirely. The code for handling \Q and \E is messy. */
3120    
3121          CHECK_RANGE:
3122          while (ptr[1] == '\\' && ptr[2] == 'E')
3123            {
3124            inescq = FALSE;
3125            ptr += 2;
3126            }
3127    
3128          oldptr = ptr;
3129    
3130          /* Remember \r or \n */
3131    
3132        if (ptr[1] == '-' && ptr[2] != ']')        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3133    
3134          /* Check for range */
3135    
3136          if (!inescq && ptr[1] == '-')
3137          {          {
3138          int d;          int d;
3139          ptr += 2;          ptr += 2;
3140            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
3141    
3142            /* If we hit \Q (not followed by \E) at this point, go into escaped
3143            mode. */
3144    
3145            while (*ptr == '\\' && ptr[1] == 'Q')
3146              {
3147              ptr += 2;
3148              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
3149              inescq = TRUE;
3150              break;
3151              }
3152    
3153            if (*ptr == 0 || (!inescq && *ptr == ']'))
3154              {
3155              ptr = oldptr;
3156              goto LONE_SINGLE_CHARACTER;
3157              }
3158    
3159  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3160          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3169  for (;; ptr++)
3169          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
3170          in such circumstances. */          in such circumstances. */
3171    
3172          if (d == '\\')          if (!inescq && d == '\\')
3173            {            {
3174            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3175            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3176    
3177            /* \b is backslash; \X is literal X; any other special means the '-'            /* \b is backspace; \X is literal X; \R is literal R; any other
3178            was literal */            special means the '-' was literal */
3179    
3180            if (d < 0)            if (d < 0)
3181              {              {
3182              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = '\b';
3183              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = 'X';
3184                else if (d == -ESC_R) d = 'R'; else
3185                {                {
3186                ptr = oldptr - 2;                ptr = oldptr;
3187                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3188                }                }
3189              }              }
3190            }            }
3191    
3192          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3193          the pre-pass. Optimize one-character ranges */          one-character ranges */
3194    
3195            if (d < c)
3196              {
3197              *errorcodeptr = ERR8;
3198              goto FAILED;
3199              }
3200    
3201          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3202    
3203            /* Remember \r or \n */
3204    
3205            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3206    
3207          /* 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
3208          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3209          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 3221  for (;; ptr++)
3221  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3222            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3223              {              {
3224              int occ, ocd;              unsigned int occ, ocd;
3225              int cc = c;              unsigned int cc = c;
3226              int origd = d;              unsigned int origd = d;
3227              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3228                {                {
3229                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3230                      ocd <= (unsigned int)d)
3231                    continue;                          /* Skip embedded ranges */
3232    
3233                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3234                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3235                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3236                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3237                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3238                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3239                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3240                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3241                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3242                  d = ocd;                  d = ocd;
3243                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3285  for (;; ptr++)
3285          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
3286          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3287    
3288          for (; c <= d; c++)          class_charcount += d - c + 1;
3289            class_lastchar = d;
3290    
3291            /* We can save a bit of time by skipping this in the pre-compile. */
3292    
3293            if (lengthptr == NULL) for (; c <= d; c++)
3294            {            {
3295            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3296            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3298  for (;; ptr++)
3298              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3299              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3300              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3301            }            }
3302    
3303          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3321  for (;; ptr++)
3321  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3322          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3323            {            {
3324            int chartype;            unsigned int othercase;
3325            int othercase;            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3326              {              {
3327              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3328              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3347  for (;; ptr++)
3347          }          }
3348        }        }
3349    
3350      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3351      loop. This "while" is the end of the "do" above. */  
3352        while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
3353    
3354        if (c == 0)                          /* Missing terminating ']' */
3355          {
3356          *errorcodeptr = ERR6;
3357          goto FAILED;
3358          }
3359    
3360    
3361    /* This code has been disabled because it would mean that \s counts as
3362    an explicit \r or \n reference, and that's not really what is wanted. Now
3363    we set the flag only if there is a literal "\r" or "\n" in the class. */
3364    
3365    #if 0
3366        /* Remember whether \r or \n are in this class */
3367    
3368        if (negate_class)
3369          {
3370          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3371          }
3372        else
3373          {
3374          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3375          }
3376    #endif
3377    
     while ((c = *(++ptr)) != ']' || inescq);  
3378    
3379      /* 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
3380      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
3381      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
3382      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3383      single-bytes only. This is an historical hangover. Maybe one day we can  
3384      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3385        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3386        operate on single-bytes only. This is an historical hangover. Maybe one day
3387        we can tidy these opcodes to handle multi-byte characters.
3388    
3389      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
3390      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 3394  for (;; ptr++)
3394      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3395    
3396  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3397      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3398            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3399  #else  #else
3400      if (class_charcount == 1)      if (class_charcount == 1)
3401  #endif  #endif
# Line 2209  for (;; ptr++) Line 3438  for (;; ptr++)
3438      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3439    
3440      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3441      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3442      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3443        the class, so any that were explicitly given as well can be ignored. If
3444        (when there are explicit characters > 255 that must be listed) there are no
3445        characters < 256, we can omit the bitmap in the actual compiled code. */
3446    
3447  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3448      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3449        {        {
3450        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3451        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3452        code += LINK_SIZE;        code += LINK_SIZE;
3453        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3454    
3455        /* 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;
3456        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3457    
3458        if (class_charcount > 0)        if (class_charcount > 0)
3459          {          {
3460          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3461            memmove(code + 32, code, class_utf8data - code);
3462          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3463          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;  
3464          }          }
3465          else code = class_utf8data;
3466    
3467        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3468    
# Line 2246  for (;; ptr++) Line 3471  for (;; ptr++)
3471        }        }
3472  #endif  #endif
3473    
3474      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, set the opcode to OP_CLASS or
3475      and copy it into the code vector. If this is the first thing in the branch,      OP_NCLASS, depending on whether the whole class was negated and whether
3476      there can be no first char setting, whatever the repeat count. Any reqbyte      there were negative specials such as \S in the class. Then copy the 32-byte
3477      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3478    
3479        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3480      if (negate_class)      if (negate_class)
3481        {        {
3482        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3483        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3484        }        }
3485      else      else
3486        {        {
       *code++ = OP_CLASS;  
3487        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3488        }        }
3489      code += 32;      code += 32;
3490      break;      break;
3491    
3492    
3493        /* ===================================================================*/
3494      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3495      has been tested above. */      has been tested above. */
3496    
# Line 2331  for (;; ptr++) Line 3558  for (;; ptr++)
3558        }        }
3559      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3560    
     /* 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;  
       }  
   
3561      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3562      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
3563      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 3591  for (;; ptr++)
3591          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3592          }          }
3593    
3594          /* If the repetition is unlimited, it pays to see if the next thing on
3595          the line is something that cannot possibly match this character. If so,
3596          automatically possessifying this item gains some performance in the case
3597          where the match fails. */
3598    
3599          if (!possessive_quantifier &&
3600              repeat_max < 0 &&
3601              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3602                options, cd))
3603            {
3604            repeat_type = 0;    /* Force greedy */
3605            possessive_quantifier = TRUE;
3606            }
3607    
3608        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3609        }        }
3610    
3611      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3612      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-
3613      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3614      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3615        currently used only for single-byte chars. */
3616    
3617      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3618        {        {
3619        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3620        c = previous[1];        c = previous[1];
3621          if (!possessive_quantifier &&
3622              repeat_max < 0 &&
3623              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3624            {
3625            repeat_type = 0;    /* Force greedy */
3626            possessive_quantifier = TRUE;
3627            }
3628        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3629        }        }
3630    
# Line 2403  for (;; ptr++) Line 3638  for (;; ptr++)
3638      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
3639        {        {
3640        uschar *oldcode;        uschar *oldcode;
3641        int prop_type;        int prop_type, prop_value;
3642        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3643        c = *previous;        c = *previous;
3644    
3645          if (!possessive_quantifier &&
3646              repeat_max < 0 &&
3647              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3648            {
3649            repeat_type = 0;    /* Force greedy */
3650            possessive_quantifier = TRUE;
3651            }
3652    
3653        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3654        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3655          previous[1] : -1;          {
3656            prop_type = previous[1];
3657            prop_value = previous[2];
3658            }
3659          else prop_type = prop_value = -1;
3660    
3661        oldcode = code;        oldcode = code;
3662        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2422  for (;; ptr++) Line 3669  for (;; ptr++)
3669        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3670        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3671    
3672        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3673    
3674        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3675    
# Line 2443  for (;; ptr++) Line 3690  for (;; ptr++)
3690          }          }
3691    
3692        /* 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
3693        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3694        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
3695        one less than the maximum. */        one less than the maximum. */
3696    
# Line 2470  for (;; ptr++) Line 3717  for (;; ptr++)
3717    
3718          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
3719          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
3720          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
3721          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
3722          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
3723    
# Line 2486  for (;; ptr++) Line 3733  for (;; ptr++)
3733  #endif  #endif
3734              {              {
3735              *code++ = c;              *code++ = c;
3736              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
3737                  {
3738                  *code++ = prop_type;
3739                  *code++ = prop_value;
3740                  }
3741              }              }
3742            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
3743            }            }
3744    
3745          /* 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
3746          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3747            UPTO is just for 1 instance, we can use QUERY instead. */
3748    
3749          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3750            {            {
# Line 2505  for (;; ptr++) Line 3757  for (;; ptr++)
3757            else            else
3758  #endif  #endif
3759            *code++ = c;            *code++ = c;
3760            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
3761                {
3762                *code++ = prop_type;
3763                *code++ = prop_value;
3764                }
3765            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3766            *code++ = OP_UPTO + repeat_type;  
3767            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3768                {
3769                *code++ = OP_QUERY + repeat_type;
3770                }
3771              else
3772                {
3773                *code++ = OP_UPTO + repeat_type;
3774                PUT2INC(code, 0, repeat_max);
3775                }
3776            }            }
3777          }          }
3778    
# Line 2524  for (;; ptr++) Line 3788  for (;; ptr++)
3788  #endif  #endif
3789        *code++ = c;        *code++ = c;
3790    
3791        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
3792        defines the required property. */        define the required property. */
3793    
3794  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3795        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
3796            {
3797            *code++ = prop_type;
3798            *code++ = prop_value;
3799            }
3800  #endif  #endif
3801        }        }
3802    
# Line 2551  for (;; ptr++) Line 3819  for (;; ptr++)
3819        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3820        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3821    
3822        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3823    
3824        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3825          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2571  for (;; ptr++) Line 3839  for (;; ptr++)
3839      /* 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
3840      cases. */      cases. */
3841    
3842      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3843               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3844        {        {
3845        register int i;        register int i;
3846        int ketoffset = 0;        int ketoffset = 0;
3847        int len = code - previous;        int len = code - previous;
3848        uschar *bralink = NULL;        uschar *bralink = NULL;
3849    
3850          /* Repeating a DEFINE group is pointless */
3851    
3852          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3853            {
3854            *errorcodeptr = ERR55;
3855            goto FAILED;
3856            }
3857    
3858        /* 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
3859        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
3860        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 2601  for (;; ptr++) Line 3877  for (;; ptr++)
3877    
3878        if (repeat_min == 0)        if (repeat_min == 0)
3879          {          {
3880          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
3881          altogether. */          output altogether, like this:
3882    
3883          if (repeat_max == 0)          ** if (repeat_max == 0)
3884            {          **   {
3885            code = previous;          **   code = previous;
3886            goto END_REPEAT;          **   goto END_REPEAT;
3887            }          **   }
3888    
3889            However, that fails when a group is referenced as a subroutine from
3890            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
3891            so that it is skipped on execution. As we don't have a list of which
3892            groups are referenced, we cannot do this selectively.
3893    
3894            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
3895            and do no more at this point. However, we do need to adjust any
3896            OP_RECURSE calls inside the group that refer to the group itself or any
3897            internal or forward referenced group, because the offset is from the
3898            start of the whole regex. Temporarily terminate the pattern while doing
3899            this. */
3900    
3901          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal group, because the offset is from the start of the whole  
         regex. Temporarily terminate the pattern while doing this. */  
   
         if (repeat_max <= 1)  
3902            {            {
3903            *code = OP_END;            *code = OP_END;
3904            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3905            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3906            code++;            code++;
3907              if (repeat_max == 0)
3908                {
3909                *previous++ = OP_SKIPZERO;
3910                goto END_REPEAT;
3911                }
3912            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
3913            }            }
3914    
# Line 2637  for (;; ptr++) Line 3924  for (;; ptr++)
3924            {            {
3925            int offset;            int offset;
3926            *code = OP_END;            *code = OP_END;
3927            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3928            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3929            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3930            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2657  for (;; ptr++) Line 3944  for (;; ptr++)
3944        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3945        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3946        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
3947        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3948          forward reference subroutine calls in the group, there will be entries on
3949          the workspace list; replicate these with an appropriate increment. */
3950    
3951        else        else
3952          {          {
3953          if (repeat_min > 1)          if (repeat_min > 1)
3954            {            {
3955            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3956            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
3957              potential integer overflow. */
3958    
3959              if (lengthptr != NULL)
3960                {
3961                int delta = (repeat_min - 1)*length_prevgroup;
3962                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3963                                                                (double)INT_MAX ||
3964                    OFLOW_MAX - *lengthptr < delta)
3965                  {
3966                  *errorcodeptr = ERR20;
3967                  goto FAILED;
3968                  }
3969                *lengthptr += delta;
3970                }
3971    
3972              /* This is compiling for real */
3973    
3974              else
3975              {              {
3976              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3977              code += len;              for (i = 1; i < repeat_min; i++)
3978                  {
3979                  uschar *hc;
3980                  uschar *this_hwm = cd->hwm;
3981                  memcpy(code, previous, len);
3982                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3983                    {
3984                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3985                    cd->hwm += LINK_SIZE;
3986                    }
3987                  save_hwm = this_hwm;
3988                  code += len;
3989                  }
3990              }              }
3991            }            }
3992    
3993          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3994          }          }
3995    
# Line 2677  for (;; ptr++) Line 3997  for (;; ptr++)
3997        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3998        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,
3999        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
4000        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
4001          replicate entries on the forward reference list. */
4002    
4003        if (repeat_max >= 0)        if (repeat_max >= 0)
4004          {          {
4005          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
4006            just adjust the length as if we had. For each repetition we must add 1
4007            to the length for BRAZERO and for all but the last repetition we must
4008            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4009            paranoid checks to avoid integer overflow. */
4010    
4011            if (lengthptr != NULL && repeat_max > 0)
4012              {
4013              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4014                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4015              if ((double)repeat_max *
4016                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4017                      > (double)INT_MAX ||
4018                  OFLOW_MAX - *lengthptr < delta)
4019                {
4020                *errorcodeptr = ERR20;
4021                goto FAILED;
4022                }
4023              *lengthptr += delta;
4024              }
4025    
4026            /* This is compiling for real */
4027    
4028            else for (i = repeat_max - 1; i >= 0; i--)
4029            {            {
4030              uschar *hc;
4031              uschar *this_hwm = cd->hwm;
4032    
4033            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
4034    
4035            /* 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 4045  for (;; ptr++)
4045              }              }
4046    
4047            memcpy(code, previous, len);            memcpy(code, previous, len);
4048              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4049                {
4050                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
4051                cd->hwm += LINK_SIZE;
4052                }
4053              save_hwm = this_hwm;
4054            code += len;            code += len;
4055            }            }
4056    
# Line 2720  for (;; ptr++) Line 4073  for (;; ptr++)
4073        /* 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
4074        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
4075        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
4076        correct offset was computed above. */        correct offset was computed above.
4077    
4078          Then, when we are doing the actual compile phase, check to see whether
4079          this group is a non-atomic one that could match an empty string. If so,
4080          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
4081          that runtime checking can be done. [This check is also applied to
4082          atomic groups at runtime, but in a different way.] */
4083    
4084        else code[-ketoffset] = OP_KETRMAX + repeat_type;        else
4085            {
4086            uschar *ketcode = code - ketoffset;
4087            uschar *bracode = ketcode - GET(ketcode, 1);
4088            *ketcode = OP_KETRMAX + repeat_type;
4089            if (lengthptr == NULL && *bracode != OP_ONCE)
4090              {
4091              uschar *scode = bracode;
4092              do
4093                {
4094                if (could_be_empty_branch(scode, ketcode, utf8))
4095                  {
4096                  *bracode += OP_SBRA - OP_BRA;
4097                  break;
4098                  }
4099                scode += GET(scode, 1);
4100                }
4101              while (*scode == OP_ALT);
4102              }
4103            }
4104        }        }
4105    
4106      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2733  for (;; ptr++) Line 4111  for (;; ptr++)
4111        goto FAILED;        goto FAILED;
4112        }        }
4113    
4114      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
4115      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
4116      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
4117      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.
4118      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
4119        but the special opcodes can optimize it a bit. The repeated item starts at
4120        tempcode, not at previous, which might be the first part of a string whose
4121        (former) last char we repeated.
4122    
4123        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4124        an 'upto' may follow. We skip over an 'exact' item, and then test the
4125        length of what remains before proceeding. */
4126    
4127      if (possessive_quantifier)      if (possessive_quantifier)
4128        {        {
4129        int len = code - tempcode;        int len;
4130        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4131        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
4132        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode] +
4133        tempcode[0] = OP_ONCE;            ((*tempcode == OP_TYPEEXACT &&
4134        *code++ = OP_KET;               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4135        PUTINC(code, 0, len);        len = code - tempcode;
4136        PUT(tempcode, 1, len);        if (len > 0) switch (*tempcode)
4137            {
4138            case OP_STAR:  *tempcode = OP_POSSTAR; break;
4139            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
4140            case OP_QUERY: *tempcode = OP_POSQUERY; break;
4141            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4142    
4143            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4144            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4145            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4146            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4147    
4148            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4149            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4150            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4151            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4152    
4153            default:
4154            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4155            code += 1 + LINK_SIZE;
4156            len += 1 + LINK_SIZE;
4157            tempcode[0] = OP_ONCE;
4158            *code++ = OP_KET;
4159            PUTINC(code, 0, len);
4160            PUT(tempcode, 1, len);
4161            break;
4162            }
4163        }        }
4164    
4165      /* 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 4172  for (;; ptr++)
4172      break;      break;
4173    
4174    
4175      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
4176      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4177      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
4178      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.  */  
4179    
4180      case '(':      case '(':
4181      newoptions = options;      newoptions = options;
4182      skipbytes = 0;      skipbytes = 0;
4183        bravalue = OP_CBRA;
4184        save_hwm = cd->hwm;
4185        reset_bracount = FALSE;
4186    
4187        /* First deal with various "verbs" that can be introduced by '*'. */
4188    
4189        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4190          {
4191          int i, namelen;
4192          const char *vn = verbnames;
4193          const uschar *name = ++ptr;
4194          previous = NULL;
4195          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4196          if (*ptr == ':')
4197            {
4198            *errorcodeptr = ERR59;   /* Not supported */
4199            goto FAILED;
4200            }
4201          if (*ptr != ')')
4202            {
4203            *errorcodeptr = ERR60;
4204            goto FAILED;
4205            }
4206          namelen = ptr - name;
4207          for (i = 0; i < verbcount; i++)
4208            {
4209            if (namelen == verbs[i].len &&
4210                strncmp((char *)name, vn, namelen) == 0)
4211              {
4212              *code = verbs[i].op;
4213              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4214              break;
4215              }
4216            vn += verbs[i].len + 1;
4217            }
4218          if (i < verbcount) continue;
4219          *errorcodeptr = ERR60;
4220          goto FAILED;
4221          }
4222    
4223      if (*(++ptr) == '?')      /* Deal with the extended parentheses; all are introduced by '?', and the
4224        appearance of any of them means that this is not a capturing group. */
4225    
4226        else if (*ptr == '?')
4227        {        {
4228        int set, unset;        int i, set, unset, namelen;
4229        int *optset;        int *optset;
4230          const uschar *name;
4231          uschar *slot;
4232    
4233        switch (*(++ptr))        switch (*(++ptr))
4234          {          {
4235          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
4236          ptr++;          ptr++;
4237          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
4238            if (*ptr == 0)
4239              {
4240              *errorcodeptr = ERR18;
4241              goto FAILED;
4242              }
4243          continue;          continue;
4244    
4245          case ':':                 /* Non-extracting bracket */  
4246            /* ------------------------------------------------------------ */
4247            case '|':                 /* Reset capture count for each branch */
4248            reset_bracount = TRUE;
4249            /* Fall through */
4250    
4251            /* ------------------------------------------------------------ */
4252            case ':':                 /* Non-capturing bracket */
4253          bravalue = OP_BRA;          bravalue = OP_BRA;
4254          ptr++;          ptr++;
4255          break;          break;
4256    
4257    
4258            /* ------------------------------------------------------------ */
4259          case '(':          case '(':
4260          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4261    
4262          /* Condition to test for recursion */          /* A condition can be an assertion, a number (referring to a numbered
4263            group), a name (referring to a named group), or 'R', referring to
4264            recursion. R<digits> and R&name are also permitted for recursion tests.
4265    
4266            There are several syntaxes for testing a named group: (?(name)) is used
4267            by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
4268    
4269            There are two unfortunate ambiguities, caused by history. (a) 'R' can
4270            be the recursive thing or the name 'R' (and similarly for 'R' followed
4271            by digits), and (b) a number could be a name that consists of digits.
4272            In both cases, we look for a name first; if not found, we try the other
4273            cases. */
4274    
4275            /* For conditions that are assertions, check the syntax, and then exit
4276            the switch. This will take control down to where bracketed groups,
4277            including assertions, are processed. */
4278    
4279            if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
4280              break;
4281    
4282            /* Most other conditions use OP_CREF (a couple change to OP_RREF
4283            below), and all need to skip 3 bytes at the start of the group. */
4284    
4285            code[1+LINK_SIZE] = OP_CREF;
4286            skipbytes = 3;
4287            refsign = -1;
4288    
4289            /* Check for a test for recursion in a named group. */
4290    
4291          if (ptr[1] == 'R')          if (ptr[1] == 'R' && ptr[2] == '&')
4292            {            {
4293            code[1+LINK_SIZE] = OP_CREF;            terminator = -1;
4294            PUT2(code, 2+LINK_SIZE, CREF_RECURSE);            ptr += 2;
4295            skipbytes = 3;            code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
           ptr += 3;  
4296            }            }
4297    
4298          /* 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
4299          if a digit follows ( then there will just be digits until ) because          syntax (?(<name>) or (?('name') */
         the syntax was checked in the first pass. */  
4300    
4301          else if ((digitab[ptr[1]] && ctype_digit) != 0)          else if (ptr[1] == '<')
4302            {            {
4303            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;  
             }  
4304            ptr++;            ptr++;
           code[1+LINK_SIZE] = OP_CREF;  
           PUT2(code, 2+LINK_SIZE, condref);  
           skipbytes = 3;  
4305            }            }
4306          /* For conditions that are assertions, we just fall through, having          else if (ptr[1] == '\'')
4307          set bravalue above. */            {
4308          break;            terminator = '\'';
4309              ptr++;
4310          case '=':                 /* Positive lookahead */            }
4311          bravalue = OP_ASSERT;          else
4312          ptr++;            {
4313          break;            terminator = 0;
4314              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4315              }
4316    
4317          case '!':                 /* Negative lookahead */          /* We now expect to read a name; any thing else is an error */
         bravalue = OP_ASSERT_NOT;  
         ptr++;  
         break;  
4318    
4319          case '<':                 /* Lookbehinds */          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
         switch (*(++ptr))  
4320            {            {
4321            case '=':               /* Positive lookbehind */            ptr += 1;  /* To get the right offset */
4322            bravalue = OP_ASSERTBACK;            *errorcodeptr = ERR28;
4323            ptr++;            goto FAILED;
4324            break;            }
4325    
4326            case '!':               /* Negative lookbehind */          /* Read the name, but also get it as a number if it's all digits */
4327            bravalue = OP_ASSERTBACK_NOT;  
4328            recno = 0;
4329            name = ++ptr;
4330            while ((cd->ctypes[*ptr] & ctype_word) != 0)
4331              {
4332              if (recno >= 0)
4333                recno = ((digitab[*ptr] & ctype_digit) != 0)?
4334                  recno * 10 + *ptr - '0' : -1;
4335            ptr++;            ptr++;
           break;  
4336            }            }
4337          break;          namelen = ptr - name;
4338    
4339          case '>':                 /* One-time brackets */          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
4340          bravalue = OP_ONCE;            {
4341          ptr++;            ptr--;      /* Error offset */
4342          break;            *errorcodeptr = ERR26;
4343              goto FAILED;
4344              }
4345    
4346          case 'C':                 /* Callout - may be followed by digits; */          /* Do no further checking in the pre-compile phase. */
4347          previous_callout = code;  /* Save for later completion */  
4348          after_manual_callout = 1; /* Skip one item before completing */          if (lengthptr != NULL) break;
4349          *code++ = OP_CALLOUT;     /* Already checked that the terminating */  
4350            {                       /* closing parenthesis is present. */          /* In the real compile we do the work of looking for the actual
4351            int n = 0;          reference. If the string started with "+" or "-" we require the rest to
4352            while ((digitab[*(++ptr)] & ctype_digit) != 0)          be digits, in which case recno will be set. */
4353              n = n * 10 + *ptr - '0';  
4354            if (n > 255)          if (refsign > 0)
4355              {
4356              if (recno <= 0)
4357              {              {
4358              *errorcodeptr = ERR38;              *errorcodeptr = ERR58;
4359              goto FAILED;              goto FAILED;
4360              }              }
4361            *code++ = n;            recno = (refsign == '-')?
4362            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */              cd->bracount - recno + 1 : recno +cd->bracount;
4363