/[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 91 by nigel, Sat Feb 24 21:41:34 2007 UTC revision 323 by ph10, Wed Mar 5 17:23:42 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) 1997-2006 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  #define NLBLOCK cd            /* The block containing newline information */  #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
# Line 54  used by pcretest. DEBUG is not defined w Line 61  used by pcretest. DEBUG is not defined w
61  #endif  #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
79    
80  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
81  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
82  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
83  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
84  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
85    so this number is very generous.
86    
87    The same workspace is used during the second, actual compile phase for
88    remembering forward references to groups so that they can be filled in at the
89    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
90    is 4 there is plenty of room. */
91    
92  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
93    
94    
95  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 73  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 97  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 116  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, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
145  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. */
146    
147  static const char *const posix_names[] = {  typedef struct verbitem {
148    "alpha", "lower", "upper",    int   len;
149    "alnum", "ascii", "blank", "cntrl", "digit", "graph",    int   op;
150    "print", "punct", "space", "word",  "xdigit" };  } verbitem;
151    
152    static const char verbnames[] =
153      "ACCEPT\0"
154      "COMMIT\0"
155      "F\0"
156      "FAIL\0"
157      "PRUNE\0"
158      "SKIP\0"
159      "THEN";
160    
161    static verbitem verbs[] = {
162      { 6, OP_ACCEPT },
163      { 6, OP_COMMIT },
164      { 1, OP_FAIL },
165      { 4, OP_FAIL },
166      { 5, OP_PRUNE },
167      { 4, OP_SKIP  },
168      { 4, OP_THEN  }
169    };
170    
171    static int verbcount = sizeof(verbs)/sizeof(verbitem);
172    
173    
174    /* Tables of names of POSIX character classes and their lengths. The names are
175    now all in a single string, to reduce the number of relocations when a shared
176    library is dynamically loaded. The list of lengths is terminated by a zero
177    length entry. The first three must be alpha, lower, upper, as this is assumed
178    for handling case independence. */
179    
180    static const char posix_names[] =
181      "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"
182      "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"
183      "word\0"   "xdigit";
184    
185  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
186    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 156  static const int posix_class_maps[] = { Line 213  static const int posix_class_maps[] = {
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 or name 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 subpatterns 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 32 characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
288    "too many named subpatterns (maximum 10,000)",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
289    /* 50 */    /* 50 */
290    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
291    "octal value is greater than \\377 (not in UTF-8 mode)"    "octal value is greater than \\377 (not in UTF-8 mode)\0"
292  };    "internal error: overran compiling workspace\0"
293      "internal error: previously-checked referenced subpattern not found\0"
294      "DEFINE group contains more than one branch\0"
295      /* 55 */
296      "repeating a DEFINE group is not allowed\0"
297      "inconsistent NEWLINE options\0"
298      "\\g is not followed by a braced name or an optionally braced non-zero number\0"
299      "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"
300      "(*VERB) with an argument is not supported\0"
301      /* 60 */
302      "(*VERB) not recognized\0"
303      "number is too big\0"
304      "subpattern name expected\0"
305      "digit expected after (?+";
306    
307    
308  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 241  For convenience, we use the same bit def Line 321  For convenience, we use the same bit def
321    
322  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
323    
324  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */
325  static const unsigned char digitab[] =  static const unsigned char digitab[] =
326    {    {
327    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 277  static const unsigned char digitab[] = Line 357  static const unsigned char digitab[] =
357    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
358    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
359    
360  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else           /* This is the "abnormal" case, for EBCDIC systems */
361  static const unsigned char digitab[] =  static const unsigned char digitab[] =
362    {    {
363    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 291  static const unsigned char digitab[] = Line 371  static const unsigned char digitab[] =
371    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
372    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
373    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
374    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
375    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
376    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
377    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 325  static const unsigned char ebcdic_charta Line 405  static const unsigned char ebcdic_charta
405    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
406    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
407    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
408    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
409    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
410    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
411    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 352  static const unsigned char ebcdic_charta Line 432  static const unsigned char ebcdic_charta
432  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
433    
434  static BOOL  static BOOL
435    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
436      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
437    
438    
439    
440    /*************************************************
441    *            Find an error text                  *
442    *************************************************/
443    
444    /* The error texts are now all in one long string, to save on relocations. As
445    some of the text is of unknown length, we can't use a table of offsets.
446    Instead, just count through the strings. This is not a performance issue
447    because it happens only when there has been a compilation error.
448    
449    Argument:   the error number
450    Returns:    pointer to the error string
451    */
452    
453    static const char *
454    find_error_text(int n)
455    {
456    const char *s = error_texts;
457    for (; n > 0; n--) while (*s++ != 0);
458    return s;
459    }
460    
461    
462  /*************************************************  /*************************************************
# Line 363  static BOOL Line 465  static BOOL
465    
466  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
467  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
468  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
469  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
470  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
471    ptr is pointing at the \. On exit, it is on the final character of the escape
472    sequence.
473    
474  Arguments:  Arguments:
475    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 376  Arguments: Line 480  Arguments:
480    
481  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
482                   negative => a special escape sequence                   negative => a special escape sequence
483                   on error, errorptr is set                   on error, errorcodeptr is set
484  */  */
485    
486  static int  static int
# Line 394  ptr--;                            /* Set Line 498  ptr--;                            /* Set
498    
499  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
500    
501  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
502  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
503  Otherwise further processing may be required. */  Otherwise further processing may be required. */
504    
505  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
506  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */
507  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
508    
509  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
510  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
511  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
512  #endif  #endif
513    
# Line 412  else if ((i = escapes[c - 0x48]) != 0) Line 516  else if ((i = escapes[c - 0x48]) != 0)
516  else  else
517    {    {
518    const uschar *oldptr;    const uschar *oldptr;
519      BOOL braced, negated;
520    
521    switch (c)    switch (c)
522      {      {
523      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
# Line 425  else Line 531  else
531      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
532      break;      break;
533    
534        /* \g must be followed by a number, either plain or braced. If positive, it
535        is an absolute backreference. If negative, it is a relative backreference.
536        This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a
537        reference to a named group. This is part of Perl's movement towards a
538        unified syntax for back references. As this is synonymous with \k{name}, we
539        fudge it up by pretending it really was \k. */
540    
541        case 'g':
542        if (ptr[1] == '{')
543          {
544          const uschar *p;
545          for (p = ptr+2; *p != 0 && *p != '}'; p++)
546            if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;
547          if (*p != 0 && *p != '}')
548            {
549            c = -ESC_k;
550            break;
551            }
552          braced = TRUE;
553          ptr++;
554          }
555        else braced = FALSE;
556    
557        if (ptr[1] == '-')
558          {
559          negated = TRUE;
560          ptr++;
561          }
562        else negated = FALSE;
563    
564        c = 0;
565        while ((digitab[ptr[1]] & ctype_digit) != 0)
566          c = c * 10 + *(++ptr) - '0';
567    
568        if (c < 0)
569          {
570          *errorcodeptr = ERR61;
571          break;
572          }
573    
574        if (c == 0 || (braced && *(++ptr) != '}'))
575          {
576          *errorcodeptr = ERR57;
577          break;
578          }
579    
580        if (negated)
581          {
582          if (c > bracount)
583            {
584            *errorcodeptr = ERR15;
585            break;
586            }
587          c = bracount - (c - 1);
588          }
589    
590        c = -(ESC_REF + c);
591        break;
592    
593      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
594      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
595      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 446  else Line 611  else
611        c -= '0';        c -= '0';
612        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
613          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
614          if (c < 0)
615            {
616            *errorcodeptr = ERR61;
617            break;
618            }
619        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
620          {          {
621          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 495  else Line 665  else
665          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == '0') continue;     /* Leading zeroes */
666          count++;          count++;
667    
668  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
669          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= 'a') cc -= 32;               /* Convert to upper case */
670          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));
671  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
672          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */
673          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));
674  #endif  #endif
# Line 522  else Line 692  else
692        {        {
693        int cc;                               /* Some compilers don't like ++ */        int cc;                               /* Some compilers don't like ++ */
694        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                        /* in initializers */
695  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
696        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= 'a') cc -= 32;              /* Convert to upper case */
697        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));
698  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
699        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= 'z') cc += 64;              /* Convert to upper case */
700        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));
701  #endif  #endif
702        }        }
703      break;      break;
704    
705      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
706        This coding is ASCII-specific, but then the whole concept of \cx is
707        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
708    
709      case 'c':      case 'c':
710      c = *(++ptr);      c = *(++ptr);
711      if (c == 0)      if (c == 0)
712        {        {
713        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
714        return 0;        break;
715        }        }
716    
717      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII coding */
     is ASCII-specific, but then the whole concept of \cx is ASCII-specific.  
     (However, an EBCDIC equivalent has now been added.) */  
   
 #if !EBCDIC    /* ASCII coding */  
718      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= 'a' && c <= 'z') c -= 32;
719      c ^= 0x40;      c ^= 0x40;
720  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
721      if (c >= 'a' && c <= 'z') c += 64;      if (c >= 'a' && c <= 'z') c += 64;
722      c ^= 0xC0;      c ^= 0xC0;
723  #endif  #endif
724      break;      break;
725    
726      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
727      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
728      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
729      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
730      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
731    
732      default:      default:
733      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 619  if (c == '{') Line 787  if (c == '{')
787      *negptr = TRUE;      *negptr = TRUE;
788      ptr++;      ptr++;
789      }      }
790    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
791      {      {
792      c = *(++ptr);      c = *(++ptr);
793      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
# Line 648  top = _pcre_utt_size; Line 816  top = _pcre_utt_size;
816  while (bot < top)  while (bot < top)
817    {    {
818    i = (bot + top) >> 1;    i = (bot + top) >> 1;
819    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
820    if (c == 0)    if (c == 0)
821      {      {
822      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 772  return p; Line 940  return p;
940    
941    
942  /*************************************************  /*************************************************
943  *     Find forward referenced named subpattern   *  *       Find forward referenced subpattern       *
944  *************************************************/  *************************************************/
945    
946  /* This function scans along a pattern looking for capturing subpatterns, and  /* This function scans along a pattern's text looking for capturing
947  counting them. If it finds a named pattern that matches the name it is given,  subpatterns, and counting them. If it finds a named pattern that matches the
948  it returns its number. This is used for forward references to named  name it is given, it returns its number. Alternatively, if the name is NULL, it
949  subpatterns. We know that if (?P< is encountered, the name will be terminated  returns when it reaches a given numbered subpattern. This is used for forward
950  by '>' because that is checked in the first pass.  references to subpatterns. We know that if (?P< is encountered, the name will
951    be terminated by '>' because that is checked in the first pass.
952    
953  Arguments:  Arguments:
954    pointer      current position in the pattern    ptr          current position in the pattern
955    count        current count of capturing parens    count        current count of capturing parens so far encountered
956    name         name to seek    name         name to seek, or NULL if seeking a numbered subpattern
957    namelen      name length    lorn         name length, or subpattern number if name is NULL
958      xmode        TRUE if we are in /x mode
959    
960  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
961  */  */
962    
963  static int  static int
964  find_named_parens(const uschar *ptr, int count, const uschar *name, int namelen)  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,
965      BOOL xmode)
966  {  {
967  const uschar *thisname;  const uschar *thisname;
968    
969  for (; *ptr != 0; ptr++)  for (; *ptr != 0; ptr++)
970    {    {
971    if (*ptr == '\\' && ptr[1] != 0) { ptr++; continue; }    int term;
972    
973      /* Skip over backslashed characters and also entire \Q...\E */
974    
975      if (*ptr == '\\')
976        {
977        if (*(++ptr) == 0) return -1;
978        if (*ptr == 'Q') for (;;)
979          {
980          while (*(++ptr) != 0 && *ptr != '\\');
981          if (*ptr == 0) return -1;
982          if (*(++ptr) == 'E') break;
983          }
984        continue;
985        }
986    
987      /* Skip over character classes */
988    
989      if (*ptr == '[')
990        {
991        while (*(++ptr) != ']')
992          {
993          if (*ptr == 0) return -1;
994          if (*ptr == '\\')
995            {
996            if (*(++ptr) == 0) return -1;
997            if (*ptr == 'Q') for (;;)
998              {
999              while (*(++ptr) != 0 && *ptr != '\\');
1000              if (*ptr == 0) return -1;
1001              if (*(++ptr) == 'E') break;
1002              }
1003            continue;
1004            }
1005          }
1006        continue;
1007        }
1008    
1009      /* Skip comments in /x mode */
1010    
1011      if (xmode && *ptr == '#')
1012        {
1013        while (*(++ptr) != 0 && *ptr != '\n');
1014        if (*ptr == 0) return -1;
1015        continue;
1016        }
1017    
1018      /* An opening parens must now be a real metacharacter */
1019    
1020    if (*ptr != '(') continue;    if (*ptr != '(') continue;
1021    if (ptr[1] != '?') { count++; continue; }    if (ptr[1] != '?' && ptr[1] != '*')
1022    if (ptr[2] == '(') { ptr += 2; continue; }      {
1023    if (ptr[2] != 'P' || ptr[3] != '<') continue;      count++;
1024        if (name == NULL && count == lorn) return count;
1025        continue;
1026        }
1027    
1028      ptr += 2;
1029      if (*ptr == 'P') ptr++;                      /* Allow optional P */
1030    
1031      /* We have to disambiguate (?<! and (?<= from (?<name> */
1032    
1033      if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&
1034           *ptr != '\'')
1035        continue;
1036    
1037    count++;    count++;
1038    ptr += 4;  
1039      if (name == NULL && count == lorn) return count;
1040      term = *ptr++;
1041      if (term == '<') term = '>';
1042    thisname = ptr;    thisname = ptr;
1043    while (*ptr != '>') ptr++;    while (*ptr != term) ptr++;
1044    if (namelen == ptr - thisname && strncmp(name, thisname, namelen) == 0)    if (name != NULL && lorn == ptr - thisname &&
1045          strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1046      return count;      return count;
1047    }    }
1048    
1049  return -1;  return -1;
1050  }  }
1051    
# Line 862  for (;;) Line 1100  for (;;)
1100    
1101      case OP_CALLOUT:      case OP_CALLOUT:
1102      case OP_CREF:      case OP_CREF:
1103      case OP_BRANUMBER:      case OP_RREF:
1104        case OP_DEF:
1105      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1106      break;      break;
1107    
# Line 907  for (;;) Line 1146  for (;;)
1146    {    {
1147    int d;    int d;
1148    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1149    switch (op)    switch (op)
1150      {      {
1151        case OP_CBRA:
1152      case OP_BRA:      case OP_BRA:
1153      case OP_ONCE:      case OP_ONCE:
1154      case OP_COND:      case OP_COND:
1155      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
1156      if (d < 0) return d;      if (d < 0) return d;
1157      branchlength += d;      branchlength += d;
1158      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 949  for (;;) Line 1187  for (;;)
1187      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1188    
1189      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1190      case OP_CREF:      case OP_CREF:
1191        case OP_RREF:
1192        case OP_DEF:
1193      case OP_OPT:      case OP_OPT:
1194      case OP_CALLOUT:      case OP_CALLOUT:
1195      case OP_SOD:      case OP_SOD:
# Line 995  for (;;) Line 1234  for (;;)
1234    
1235      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1236      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1237        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1238      cc += 4;      cc += 4;
1239      break;      break;
1240    
# Line 1094  for (;;) Line 1334  for (;;)
1334    
1335    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1336    
1337    /* Handle bracketed group */    /* Handle capturing bracket */
1338    
1339    else if (c > OP_BRA)    else if (c == OP_CBRA)
1340      {      {
1341      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1342      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1343      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1344      }      }
1345    
1346    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1347    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1348    The length in the table is a minimum, so we have to scan along to skip the    two bytes of parameters. */
   extra bytes. All opcodes are less than 128, so we can use relatively  
   efficient code. */  
1349    
1350    else    else
1351      {      {
1352        switch(c)
1353          {
1354          case OP_TYPESTAR:
1355          case OP_TYPEMINSTAR:
1356          case OP_TYPEPLUS:
1357          case OP_TYPEMINPLUS:
1358          case OP_TYPEQUERY:
1359          case OP_TYPEMINQUERY:
1360          case OP_TYPEPOSSTAR:
1361          case OP_TYPEPOSPLUS:
1362          case OP_TYPEPOSQUERY:
1363          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1364          break;
1365    
1366          case OP_TYPEUPTO:
1367          case OP_TYPEMINUPTO:
1368          case OP_TYPEEXACT:
1369          case OP_TYPEPOSUPTO:
1370          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1371          break;
1372          }
1373    
1374        /* Add in the fixed length from the table */
1375    
1376      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1377    
1378      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1379      a multi-byte character. The length in the table is a minimum, so we have to
1380      arrange to skip the extra bytes. */
1381    
1382    #ifdef SUPPORT_UTF8
1383      if (utf8) switch(c)      if (utf8) switch(c)
1384        {        {
1385        case OP_CHAR:        case OP_CHAR:
# Line 1120  for (;;) Line 1387  for (;;)
1387        case OP_EXACT:        case OP_EXACT:
1388        case OP_UPTO:        case OP_UPTO:
1389        case OP_MINUPTO:        case OP_MINUPTO:
1390          case OP_POSUPTO:
1391        case OP_STAR:        case OP_STAR:
1392        case OP_MINSTAR:        case OP_MINSTAR:
1393          case OP_POSSTAR:
1394        case OP_PLUS:        case OP_PLUS:
1395        case OP_MINPLUS:        case OP_MINPLUS:
1396          case OP_POSPLUS:
1397        case OP_QUERY:        case OP_QUERY:
1398        case OP_MINQUERY:        case OP_MINQUERY:
1399        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1400          if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1401        break;        break;
1402        }        }
1403    #endif
1404      }      }
1405    }    }
1406  }  }
# Line 1164  for (;;) Line 1436  for (;;)
1436    
1437    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1438    
1439    /* All bracketed groups have the same length. */    /* Otherwise, we can get the item's length from the table, except that for
1440      repeated character types, we have to test for \p and \P, which have an extra
1441      two bytes of parameters. */
1442    
1443    else if (c > OP_BRA)    else
1444      {      {
1445      code += _pcre_OP_lengths[OP_BRA];      switch(c)
1446      }        {
1447          case OP_TYPESTAR:
1448          case OP_TYPEMINSTAR:
1449          case OP_TYPEPLUS:
1450          case OP_TYPEMINPLUS:
1451          case OP_TYPEQUERY:
1452          case OP_TYPEMINQUERY:
1453          case OP_TYPEPOSSTAR:
1454          case OP_TYPEPOSPLUS:
1455          case OP_TYPEPOSQUERY:
1456          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1457          break;
1458    
1459    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes        case OP_TYPEPOSUPTO:
1460    that are followed by a character may be followed by a multi-byte character.        case OP_TYPEUPTO:
1461    The length in the table is a minimum, so we have to scan along to skip the        case OP_TYPEMINUPTO:
1462    extra bytes. All opcodes are less than 128, so we can use relatively        case OP_TYPEEXACT:
1463    efficient code. */        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1464          break;
1465          }
1466    
1467        /* Add in the fixed length from the table */
1468    
   else  
     {  
1469      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1470    
1471        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1472        by a multi-byte character. The length in the table is a minimum, so we have
1473        to arrange to skip the extra bytes. */
1474    
1475    #ifdef SUPPORT_UTF8
1476      if (utf8) switch(c)      if (utf8) switch(c)
1477        {        {
1478        case OP_CHAR:        case OP_CHAR:
# Line 1187  for (;;) Line 1480  for (;;)
1480        case OP_EXACT:        case OP_EXACT:
1481        case OP_UPTO:        case OP_UPTO:
1482        case OP_MINUPTO:        case OP_MINUPTO:
1483          case OP_POSUPTO:
1484        case OP_STAR:        case OP_STAR:
1485        case OP_MINSTAR:        case OP_MINSTAR:
1486          case OP_POSSTAR:
1487        case OP_PLUS:        case OP_PLUS:
1488        case OP_MINPLUS:        case OP_MINPLUS:
1489          case OP_POSPLUS:
1490        case OP_QUERY:        case OP_QUERY:
1491        case OP_MINQUERY:        case OP_MINQUERY:
1492        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1493          if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1494        break;        break;
1495        }        }
1496    #endif
1497      }      }
1498    }    }
1499  }  }
# Line 1207  for (;;) Line 1505  for (;;)
1505  *************************************************/  *************************************************/
1506    
1507  /* This function scans through a branch of a compiled pattern to see whether it  /* This function scans through a branch of a compiled pattern to see whether it
1508  can match the empty string or not. It is called only from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1509  below. Note that first_significant_code() skips over assertions. If we hit an  below and from compile_branch() when checking for an unlimited repeat of a
1510  unclosed bracket, we return "empty" - this means we've struck an inner bracket  group that can match nothing. Note that first_significant_code() skips over
1511  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1512    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1513    bracket whose current branch will already have been scanned.
1514    
1515  Arguments:  Arguments:
1516    code        points to start of search    code        points to start of search
# Line 1224  static BOOL Line 1524  static BOOL
1524  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)
1525  {  {
1526  register int c;  register int c;
1527  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
1528       code < endcode;       code < endcode;
1529       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1530    {    {
# Line 1232  for (code = first_significant_code(code Line 1532  for (code = first_significant_code(code
1532    
1533    c = *code;    c = *code;
1534    
1535    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1536      first_significant_code() with a TRUE final argument. */
1537    
1538      if (c == OP_ASSERT)
1539        {
1540        do code += GET(code, 1); while (*code == OP_ALT);
1541        c = *code;
1542        continue;
1543        }
1544    
1545      /* Groups with zero repeats can of course be empty; skip them. */
1546    
1547      if (c == OP_BRAZERO || c == OP_BRAMINZERO)
1548        {
1549        code += _pcre_OP_lengths[c];
1550        do code += GET(code, 1); while (*code == OP_ALT);
1551        c = *code;
1552        continue;
1553        }
1554    
1555      /* For other groups, scan the branches. */
1556    
1557      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1558      {      {
1559      BOOL empty_branch;      BOOL empty_branch;
1560      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1248  for (code = first_significant_code(code Line 1570  for (code = first_significant_code(code
1570        }        }
1571      while (*code == OP_ALT);      while (*code == OP_ALT);
1572      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
     code += 1 + LINK_SIZE;  
1573      c = *code;      c = *code;
1574        continue;
1575      }      }
1576    
1577    else switch (c)    /* Handle the other opcodes */
1578    
1579      switch (c)
1580      {      {
1581      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1582        cannot be represented just by a bit map. This includes negated single
1583        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1584        actual length is stored in the compiled code, so we must update "code"
1585        here. */
1586    
1587  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1588      case OP_XCLASS:      case OP_XCLASS:
1589      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1590      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1591  #endif  #endif
1592    
# Line 1308  for (code = first_significant_code(code Line 1636  for (code = first_significant_code(code
1636      case OP_NOT:      case OP_NOT:
1637      case OP_PLUS:      case OP_PLUS:
1638      case OP_MINPLUS:      case OP_MINPLUS:
1639        case OP_POSPLUS:
1640      case OP_EXACT:      case OP_EXACT:
1641      case OP_NOTPLUS:      case OP_NOTPLUS:
1642      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
1643        case OP_NOTPOSPLUS:
1644      case OP_NOTEXACT:      case OP_NOTEXACT:
1645      case OP_TYPEPLUS:      case OP_TYPEPLUS:
1646      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
1647        case OP_TYPEPOSPLUS:
1648      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1649      return FALSE;      return FALSE;
1650    
1651        /* These are going to continue, as they may be empty, but we have to
1652        fudge the length for the \p and \P cases. */
1653    
1654        case OP_TYPESTAR:
1655        case OP_TYPEMINSTAR:
1656        case OP_TYPEPOSSTAR:
1657        case OP_TYPEQUERY:
1658        case OP_TYPEMINQUERY:
1659        case OP_TYPEPOSQUERY:
1660        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1661        break;
1662    
1663        /* Same for these */
1664    
1665        case OP_TYPEUPTO:
1666        case OP_TYPEMINUPTO:
1667        case OP_TYPEPOSUPTO:
1668        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1669        break;
1670    
1671      /* End of branch */      /* End of branch */
1672    
1673      case OP_KET:      case OP_KET:
# Line 1325  for (code = first_significant_code(code Line 1676  for (code = first_significant_code(code
1676      case OP_ALT:      case OP_ALT:
1677      return TRUE;      return TRUE;
1678    
1679      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
1680      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
1681    
1682  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1683      case OP_STAR:      case OP_STAR:
1684      case OP_MINSTAR:      case OP_MINSTAR:
1685        case OP_POSSTAR:
1686      case OP_QUERY:      case OP_QUERY:
1687      case OP_MINQUERY:      case OP_MINQUERY:
1688        case OP_POSQUERY:
1689      case OP_UPTO:      case OP_UPTO:
1690      case OP_MINUPTO:      case OP_MINUPTO:
1691        case OP_POSUPTO:
1692      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;
1693      break;      break;
1694  #endif  #endif
# Line 1383  return TRUE; Line 1737  return TRUE;
1737  *************************************************/  *************************************************/
1738    
1739  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1740  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
1741  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1742  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1743    
1744    Originally, this function only recognized a sequence of letters between the
1745    terminators, but it seems that Perl recognizes any sequence of characters,
1746    though of course unknown POSIX names are subsequently rejected. Perl gives an
1747    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1748    didn't consider this to be a POSIX class. Likewise for [:1234:].
1749    
1750    The problem in trying to be exactly like Perl is in the handling of escapes. We
1751    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1752    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1753    below handles the special case of \], but does not try to do any other escape
1754    processing. This makes it different from Perl for cases such as [:l\ower:]
1755    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1756    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1757    I think.
1758    
1759  Argument:  Arguments:
1760    ptr      pointer to the initial [    ptr      pointer to the initial [
1761    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1762    
1763  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1764  */  */
1765    
1766  static BOOL  static BOOL
1767  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1768  {  {
1769  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1770  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1771  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1772    {    {
1773    *endptr = ptr;    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1774    return TRUE;      {
1775        if (*ptr == ']') return FALSE;
1776        if (*ptr == terminator && ptr[1] == ']')
1777          {
1778          *endptr = ptr;
1779          return TRUE;
1780          }
1781        }
1782    }    }
1783  return FALSE;  return FALSE;
1784  }  }
# Line 1430  Returns:     a value representing the na Line 1803  Returns:     a value representing the na
1803  static int  static int
1804  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1805  {  {
1806    const char *pn = posix_names;
1807  register int yield = 0;  register int yield = 0;
1808  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1809    {    {
1810    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1811      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1812      pn += posix_name_lengths[yield] + 1;
1813    yield++;    yield++;
1814    }    }
1815  return -1;  return -1;
# Line 1452  earlier groups that are outside the curr Line 1827  earlier groups that are outside the curr
1827  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before
1828  it, after it has been compiled. This means that any OP_RECURSE items within it  it, after it has been compiled. This means that any OP_RECURSE items within it
1829  that refer to the group itself or any contained groups have to have their  that refer to the group itself or any contained groups have to have their
1830  offsets adjusted. That is the job of this function. Before it is called, the  offsets adjusted. That one of the jobs of this function. Before it is called,
1831  partially compiled regex must be temporarily terminated with OP_END.  the partially compiled regex must be temporarily terminated with OP_END.
1832    
1833    This function has been extended with the possibility of forward references for
1834    recursions and subroutine calls. It must also check the list of such references
1835    for the group we are dealing with. If it finds that one of the recursions in
1836    the current group is on this list, it adjusts the offset in the list, not the
1837    value in the reference (which is a group number).
1838    
1839  Arguments:  Arguments:
1840    group      points to the start of the group    group      points to the start of the group
1841    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
1842    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
1843    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
1844      save_hwm   the hwm forward reference pointer at the start of the group
1845    
1846  Returns:     nothing  Returns:     nothing
1847  */  */
1848    
1849  static void  static void
1850  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
1851      uschar *save_hwm)
1852  {  {
1853  uschar *ptr = group;  uschar *ptr = group;
1854    
1855  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1856    {    {
1857    int offset = GET(ptr, 1);    int offset;
1858    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
1859    
1860      /* See if this recursion is on the forward reference list. If so, adjust the
1861      reference. */
1862    
1863      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1864        {
1865        offset = GET(hc, 0);
1866        if (cd->start_code + offset == ptr + 1)
1867          {
1868          PUT(hc, 0, offset + adjust);
1869          break;
1870          }
1871        }
1872    
1873      /* Otherwise, adjust the recursion offset if it's after the start of this
1874      group. */
1875    
1876      if (hc >= cd->hwm)
1877        {
1878        offset = GET(ptr, 1);
1879        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
1880        }
1881    
1882    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
1883    }    }
1884  }  }
# Line 1550  Yield:        TRUE when range returned; Line 1957  Yield:        TRUE when range returned;
1957  */  */
1958    
1959  static BOOL  static BOOL
1960  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
1961      unsigned int *odptr)
1962  {  {
1963  int c, othercase, next;  unsigned int c, othercase, next;
1964    
1965  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
1966    { if ((othercase = _pcre_ucp_othercase(c)) >= 0) break; }    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }
1967    
1968  if (c > d) return FALSE;  if (c > d) return FALSE;
1969    
# Line 1576  return TRUE; Line 1984  return TRUE;
1984  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
1985    
1986    
1987    
1988  /*************************************************  /*************************************************
1989  *           Compile one branch                   *  *     Check if auto-possessifying is possible    *
1990  *************************************************/  *************************************************/
1991    
1992  /* 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
1993  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
1994  bits.  sense to automatically possessify the repeated item.
1995    
1996  Arguments:  Arguments:
1997    optionsptr     pointer to the option bits    op_code       the repeated op code
1998    brackets       points to number of extracting brackets used    this          data for this item, depends on the opcode
1999    codeptr        points to the pointer to the current code point    utf8          TRUE in UTF-8 mode
2000    ptrptr         points to the current pattern pointer    utf8_char     used for utf8 character bytes, NULL if not relevant
2001    errorcodeptr   points to error code variable    ptr           next character in pattern
2002    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    options       options bits
2003    reqbyteptr     set to the last literal character required, else < 0    cd            contains pointers to tables etc.
   bcptr          points to current branch chain  
   cd             contains pointers to tables etc.  
2004    
2005  Returns:         TRUE on success  Returns:        TRUE if possessifying is wanted
                  FALSE, with *errorcodeptr set non-zero on error  
2006  */  */
2007    
2008  static BOOL  static BOOL
2009  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2010    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    const uschar *ptr, int options, compile_data *cd)
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
2011  {  {
2012  int repeat_type, op_type;  int next;
2013  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  
2014  int bravalue = 0;  /* Skip whitespace and comments in extended mode */
2015  int greedy_default, greedy_non_default;  
2016  int firstbyte, reqbyte;  if ((options & PCRE_EXTENDED) != 0)
2017  int zeroreqbyte, zerofirstbyte;    {
2018  int req_caseopt, reqvary, tempreqvary;    for (;;)
2019  int options = *optionsptr;      {
2020  int after_manual_callout = 0;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2021  register int c;      if (*ptr == '#')
2022  register uschar *code = *codeptr;        {
2023  uschar *tempcode;        while (*(++ptr) != 0)
2024  BOOL inescq = FALSE;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2025  BOOL groupsetfirstbyte = FALSE;        }
2026  const uschar *ptr = *ptrptr;      else break;
2027  const uschar *tempptr;      }
2028  uschar *previous = NULL;    }
2029  uschar *previous_callout = NULL;  
2030  uschar classbits[32];  /* If the next item is one that we can handle, get its value. A non-negative
2031    value is a character, a negative value is an escape value. */
2032    
2033    if (*ptr == '\\')
2034      {
2035      int temperrorcode = 0;
2036      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2037      if (temperrorcode != 0) return FALSE;
2038      ptr++;    /* Point after the escape sequence */
2039      }
2040    
2041    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2042      {
2043  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2044  BOOL class_utf8;    if (utf8) { GETCHARINC(next, ptr); } else
 BOOL utf8 = (options & PCRE_UTF8) != 0;  
 uschar *class_utf8data;  
 uschar utf8_char[6];  
 #else  
 BOOL utf8 = FALSE;  
2045  #endif  #endif
2046      next = *ptr++;
2047      }
2048    
2049  /* Set up the default and non-default settings for greediness */  else return FALSE;
2050    
2051  greedy_default = ((options & PCRE_UNGREEDY) != 0);  /* Skip whitespace and comments in extended mode */
 greedy_non_default = greedy_default ^ 1;  
2052    
2053  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  if ((options & PCRE_EXTENDED) != 0)
2054  matching encountered yet". It gets changed to REQ_NONE if we hit something that    {
2055  matches a non-fixed char first char; reqbyte just remains unset if we never    for (;;)
2056  find one.      {
2057        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2058        if (*ptr == '#')
2059          {
2060          while (*(++ptr) != 0)
2061            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2062          }
2063        else break;
2064        }
2065      }
2066    
2067  When we hit a repeat whose minimum is zero, we may have to adjust these values  /* If the next thing is itself optional, we have to give up. */
 to take the zero repeat into account. This is implemented by setting them to  
 zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  
 item types that can be repeated set these backoff variables appropriately. */  
2068    
2069  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)
2070      return FALSE;
2071    
2072  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* Now compare the next item with the previous opcode. If the previous is a
2073  according to the current setting of the caseless flag. REQ_CASELESS is a bit  positive single character match, "item" either contains the character or, if
2074  value > 255. It is added into the firstbyte or reqbyte variables to record the  "item" is greater than 127 in utf8 mode, the character's bytes are in
2075  case status of the value. This is used only for ASCII characters. */  utf8_char. */
2076    
 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  
2077    
2078  /* Switch on next character until the end of the branch */  /* Handle cases when the next item is a character. */
2079    
2080  for (;; ptr++)  if (next >= 0) switch(op_code)
2081    {    {
2082    BOOL negate_class;    case OP_CHAR:
2083    BOOL possessive_quantifier;  #ifdef SUPPORT_UTF8
2084    BOOL is_quantifier;    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2085    int class_charcount;  #endif
2086    int class_lastchar;    return item != next;
2087    int newoptions;  
2088    int recno;    /* For CHARNC (caseless character) we must check the other case. If we have
2089      Unicode property support, we can use it to test the other case of
2090      high-valued characters. */
2091    
2092      case OP_CHARNC:
2093    #ifdef SUPPORT_UTF8
2094      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2095    #endif
2096      if (item == next) return FALSE;
2097    #ifdef SUPPORT_UTF8
2098      if (utf8)
2099        {
2100        unsigned int othercase;
2101        if (next < 128) othercase = cd->fcc[next]; else
2102    #ifdef SUPPORT_UCP
2103        othercase = _pcre_ucp_othercase((unsigned int)next);
2104    #else
2105        othercase = NOTACHAR;
2106    #endif
2107        return (unsigned int)item != othercase;
2108        }
2109      else
2110    #endif  /* SUPPORT_UTF8 */
2111      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2112    
2113      /* For OP_NOT, "item" must be a single-byte character. */
2114    
2115      case OP_NOT:
2116      if (item == next) return TRUE;
2117      if ((options & PCRE_CASELESS) == 0) return FALSE;
2118    #ifdef SUPPORT_UTF8
2119      if (utf8)
2120        {
2121        unsigned int othercase;
2122        if (next < 128) othercase = cd->fcc[next]; else
2123    #ifdef SUPPORT_UCP
2124        othercase = _pcre_ucp_othercase(next);
2125    #else
2126        othercase = NOTACHAR;
2127    #endif
2128        return (unsigned int)item == othercase;
2129        }
2130      else
2131    #endif  /* SUPPORT_UTF8 */
2132      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2133    
2134      case OP_DIGIT:
2135      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2136    
2137      case OP_NOT_DIGIT:
2138      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2139    
2140      case OP_WHITESPACE:
2141      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2142    
2143      case OP_NOT_WHITESPACE:
2144      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2145    
2146      case OP_WORDCHAR:
2147      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2148    
2149      case OP_NOT_WORDCHAR:
2150      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2151    
2152      case OP_HSPACE:
2153      case OP_NOT_HSPACE:
2154      switch(next)
2155        {
2156        case 0x09:
2157        case 0x20:
2158        case 0xa0:
2159        case 0x1680:
2160        case 0x180e:
2161        case 0x2000:
2162        case 0x2001:
2163        case 0x2002:
2164        case 0x2003:
2165        case 0x2004:
2166        case 0x2005:
2167        case 0x2006:
2168        case 0x2007:
2169        case 0x2008:
2170        case 0x2009:
2171        case 0x200A:
2172        case 0x202f:
2173        case 0x205f:
2174        case 0x3000:
2175        return op_code != OP_HSPACE;
2176        default:
2177        return op_code == OP_HSPACE;
2178        }
2179    
2180      case OP_VSPACE:
2181      case OP_NOT_VSPACE:
2182      switch(next)
2183        {
2184        case 0x0a:
2185        case 0x0b:
2186        case 0x0c:
2187        case 0x0d:
2188        case 0x85:
2189        case 0x2028:
2190        case 0x2029:
2191        return op_code != OP_VSPACE;
2192        default:
2193        return op_code == OP_VSPACE;
2194        }
2195    
2196      default:
2197      return FALSE;
2198      }
2199    
2200    
2201    /* Handle the case when the next item is \d, \s, etc. */
2202    
2203    switch(op_code)
2204      {
2205      case OP_CHAR:
2206      case OP_CHARNC:
2207    #ifdef SUPPORT_UTF8
2208      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2209    #endif
2210      switch(-next)
2211        {
2212        case ESC_d:
2213        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2214    
2215        case ESC_D:
2216        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2217    
2218        case ESC_s:
2219        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2220    
2221        case ESC_S:
2222        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2223    
2224        case ESC_w:
2225        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2226    
2227        case ESC_W:
2228        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2229    
2230        case ESC_h:
2231        case ESC_H:
2232        switch(item)
2233          {
2234          case 0x09:
2235          case 0x20:
2236          case 0xa0:
2237          case 0x1680:
2238          case 0x180e:
2239          case 0x2000:
2240          case 0x2001:
2241          case 0x2002:
2242          case 0x2003:
2243          case 0x2004:
2244          case 0x2005:
2245          case 0x2006:
2246          case 0x2007:
2247          case 0x2008:
2248          case 0x2009:
2249          case 0x200A:
2250          case 0x202f:
2251          case 0x205f:
2252          case 0x3000:
2253          return -next != ESC_h;
2254          default:
2255          return -next == ESC_h;
2256          }
2257    
2258        case ESC_v:
2259        case ESC_V:
2260        switch(item)
2261          {
2262          case 0x0a:
2263          case 0x0b:
2264          case 0x0c:
2265          case 0x0d:
2266          case 0x85:
2267          case 0x2028:
2268          case 0x2029:
2269          return -next != ESC_v;
2270          default:
2271          return -next == ESC_v;
2272          }
2273    
2274        default:
2275        return FALSE;
2276        }
2277    
2278      case OP_DIGIT:
2279      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2280             next == -ESC_h || next == -ESC_v;
2281    
2282      case OP_NOT_DIGIT:
2283      return next == -ESC_d;
2284    
2285      case OP_WHITESPACE:
2286      return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2287    
2288      case OP_NOT_WHITESPACE:
2289      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2290    
2291      case OP_HSPACE:
2292      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2293    
2294      case OP_NOT_HSPACE:
2295      return next == -ESC_h;
2296    
2297      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2298      case OP_VSPACE:
2299      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2300    
2301      case OP_NOT_VSPACE:
2302      return next == -ESC_v;
2303    
2304      case OP_WORDCHAR:
2305      return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2306    
2307      case OP_NOT_WORDCHAR:
2308      return next == -ESC_w || next == -ESC_d;
2309    
2310      default:
2311      return FALSE;
2312      }
2313    
2314    /* Control does not reach here */
2315    }
2316    
2317    
2318    
2319    /*************************************************
2320    *           Compile one branch                   *
2321    *************************************************/
2322    
2323    /* Scan the pattern, compiling it into the a vector. If the options are
2324    changed during the branch, the pointer is used to change the external options
2325    bits. This function is used during the pre-compile phase when we are trying
2326    to find out the amount of memory needed, as well as during the real compile
2327    phase. The value of lengthptr distinguishes the two phases.
2328    
2329    Arguments:
2330      optionsptr     pointer to the option bits
2331      codeptr        points to the pointer to the current code point
2332      ptrptr         points to the current pattern pointer
2333      errorcodeptr   points to error code variable
2334      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2335      reqbyteptr     set to the last literal character required, else < 0
2336      bcptr          points to current branch chain
2337      cd             contains pointers to tables etc.
2338      lengthptr      NULL during the real compile phase
2339                     points to length accumulator during pre-compile phase
2340    
2341    Returns:         TRUE on success
2342                     FALSE, with *errorcodeptr set non-zero on error
2343    */
2344    
2345    static BOOL
2346    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2347      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2348      compile_data *cd, int *lengthptr)
2349    {
2350    int repeat_type, op_type;
2351    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2352    int bravalue = 0;
2353    int greedy_default, greedy_non_default;
2354    int firstbyte, reqbyte;
2355    int zeroreqbyte, zerofirstbyte;
2356    int req_caseopt, reqvary, tempreqvary;
2357    int options = *optionsptr;
2358    int after_manual_callout = 0;
2359    int length_prevgroup = 0;
2360    register int c;
2361    register uschar *code = *codeptr;
2362    uschar *last_code = code;
2363    uschar *orig_code = code;
2364    uschar *tempcode;
2365    BOOL inescq = FALSE;
2366    BOOL groupsetfirstbyte = FALSE;
2367    const uschar *ptr = *ptrptr;
2368    const uschar *tempptr;
2369    uschar *previous = NULL;
2370    uschar *previous_callout = NULL;
2371    uschar *save_hwm = NULL;
2372    uschar classbits[32];
2373    
2374    #ifdef SUPPORT_UTF8
2375    BOOL class_utf8;
2376    BOOL utf8 = (options & PCRE_UTF8) != 0;
2377    uschar *class_utf8data;
2378    uschar *class_utf8data_base;
2379    uschar utf8_char[6];
2380    #else
2381    BOOL utf8 = FALSE;
2382    uschar *utf8_char = NULL;
2383    #endif
2384    
2385    #ifdef DEBUG
2386    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2387    #endif
2388    
2389    /* Set up the default and non-default settings for greediness */
2390    
2391    greedy_default = ((options & PCRE_UNGREEDY) != 0);
2392    greedy_non_default = greedy_default ^ 1;
2393    
2394    /* Initialize no first byte, no required byte. REQ_UNSET means "no char
2395    matching encountered yet". It gets changed to REQ_NONE if we hit something that
2396    matches a non-fixed char first char; reqbyte just remains unset if we never
2397    find one.
2398    
2399    When we hit a repeat whose minimum is zero, we may have to adjust these values
2400    to take the zero repeat into account. This is implemented by setting them to
2401    zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
2402    item types that can be repeated set these backoff variables appropriately. */
2403    
2404    firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
2405    
2406    /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
2407    according to the current setting of the caseless flag. REQ_CASELESS is a bit
2408    value > 255. It is added into the firstbyte or reqbyte variables to record the
2409    case status of the value. This is used only for ASCII characters. */
2410    
2411    req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
2412    
2413    /* Switch on next character until the end of the branch */
2414    
2415    for (;; ptr++)
2416      {
2417      BOOL negate_class;
2418      BOOL should_flip_negation;
2419      BOOL possessive_quantifier;
2420      BOOL is_quantifier;
2421      BOOL is_recurse;
2422      BOOL reset_bracount;
2423      int class_charcount;
2424      int class_lastchar;
2425      int newoptions;
2426      int recno;
2427      int refsign;
2428    int skipbytes;    int skipbytes;
2429    int subreqbyte;    int subreqbyte;
2430    int subfirstbyte;    int subfirstbyte;
2431      int terminator;
2432    int mclength;    int mclength;
2433    uschar mcbuffer[8];    uschar mcbuffer[8];
2434    
2435    /* Next byte in the pattern */    /* Get next byte in the pattern */
2436    
2437    c = *ptr;    c = *ptr;
2438    
2439      /* If we are in the pre-compile phase, accumulate the length used for the
2440      previous cycle of this loop. */
2441    
2442      if (lengthptr != NULL)
2443        {
2444    #ifdef DEBUG
2445        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2446    #endif
2447        if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
2448          {
2449          *errorcodeptr = ERR52;
2450          goto FAILED;
2451          }
2452    
2453        /* There is at least one situation where code goes backwards: this is the
2454        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2455        the class is simply eliminated. However, it is created first, so we have to
2456        allow memory for it. Therefore, don't ever reduce the length at this point.
2457        */
2458    
2459        if (code < last_code) code = last_code;
2460    
2461        /* Paranoid check for integer overflow */
2462    
2463        if (OFLOW_MAX - *lengthptr < code - last_code)
2464          {
2465          *errorcodeptr = ERR20;
2466          goto FAILED;
2467          }
2468    
2469        *lengthptr += code - last_code;
2470        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2471    
2472        /* If "previous" is set and it is not at the start of the work space, move
2473        it back to there, in order to avoid filling up the work space. Otherwise,
2474        if "previous" is NULL, reset the current code pointer to the start. */
2475    
2476        if (previous != NULL)
2477          {
2478          if (previous > orig_code)
2479            {
2480            memmove(orig_code, previous, code - previous);
2481            code -= previous - orig_code;
2482            previous = orig_code;
2483            }
2484          }
2485        else code = orig_code;
2486    
2487        /* Remember where this code item starts so we can pick up the length
2488        next time round. */
2489    
2490        last_code = code;
2491        }
2492    
2493      /* In the real compile phase, just check the workspace used by the forward
2494      reference list. */
2495    
2496      else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)
2497        {
2498        *errorcodeptr = ERR52;
2499        goto FAILED;
2500        }
2501    
2502    /* 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 */
2503    
2504    if (inescq && c != 0)    if (inescq && c != 0)
# Line 1692  for (;; ptr++) Line 2513  for (;; ptr++)
2513        {        {
2514        if (previous_callout != NULL)        if (previous_callout != NULL)
2515          {          {
2516          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2517              complete_callout(previous_callout, ptr, cd);
2518          previous_callout = NULL;          previous_callout = NULL;
2519          }          }
2520        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1713  for (;; ptr++) Line 2535  for (;; ptr++)
2535    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2536         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2537      {      {
2538      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2539          complete_callout(previous_callout, ptr, cd);
2540      previous_callout = NULL;      previous_callout = NULL;
2541      }      }
2542    
# Line 1724  for (;; ptr++) Line 2547  for (;; ptr++)
2547      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2548      if (c == '#')      if (c == '#')
2549        {        {
2550        while (*(++ptr) != 0) if (IS_NEWLINE(ptr)) break;        while (*(++ptr) != 0)
       if (*ptr != 0)  
2551          {          {
2552          ptr += cd->nllen - 1;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
         continue;  
2553          }          }
2554          if (*ptr != 0) continue;
2555    
2556        /* Else fall through to handle end of string */        /* Else fall through to handle end of string */
2557        c = 0;        c = 0;
2558        }        }
# Line 1745  for (;; ptr++) Line 2568  for (;; ptr++)
2568    
2569    switch(c)    switch(c)
2570      {      {
2571      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
2572        case 0:                        /* The branch terminates at string end */
2573      case 0:      case '|':                      /* or | or ) */
     case '|':  
2574      case ')':      case ')':
2575      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2576      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2577      *codeptr = code;      *codeptr = code;
2578      *ptrptr = ptr;      *ptrptr = ptr;
2579        if (lengthptr != NULL)
2580          {
2581          if (OFLOW_MAX - *lengthptr < code - last_code)
2582            {
2583            *errorcodeptr = ERR20;
2584            goto FAILED;
2585            }
2586          *lengthptr += code - last_code;   /* To include callout length */
2587          DPRINTF((">> end branch\n"));
2588          }
2589      return TRUE;      return TRUE;
2590    
2591    
2592        /* ===================================================================*/
2593      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2594      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2595    
# Line 1784  for (;; ptr++) Line 2618  for (;; ptr++)
2618      *code++ = OP_ANY;      *code++ = OP_ANY;
2619      break;      break;
2620    
2621    
2622        /* ===================================================================*/
2623      /* Character classes. If the included characters are all < 256, we build a      /* Character classes. If the included characters are all < 256, we build a
2624      32-byte bitmap of the permitted characters, except in the special case      32-byte bitmap of the permitted characters, except in the special case
2625      where there is only one such character. For negated classes, we build the      where there is only one such character. For negated classes, we build the
# Line 1803  for (;; ptr++) Line 2639  for (;; ptr++)
2639      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. */
2640    
2641      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2642          check_posix_syntax(ptr, &tempptr, cd))          check_posix_syntax(ptr, &tempptr))
2643        {        {
2644        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;
2645        goto FAILED;        goto FAILED;
2646        }        }
2647    
2648      /* 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,
2649        if the first few characters (either before or after ^) are \Q\E or \E we
2650        skip them too. This makes for compatibility with Perl. */
2651    
2652      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2653        for (;;)
2654        {        {
       negate_class = TRUE;  
2655        c = *(++ptr);        c = *(++ptr);
2656          if (c == '\\')
2657            {
2658            if (ptr[1] == 'E') ptr++;
2659              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2660                else break;
2661            }
2662          else if (!negate_class && c == '^')
2663            negate_class = TRUE;
2664          else break;
2665        }        }
2666      else  
2667        {      /* If a class contains a negative special such as \S, we need to flip the
2668        negate_class = FALSE;      negation flag at the end, so that support for characters > 255 works
2669        }      correctly (they are all included in the class). */
2670    
2671        should_flip_negation = FALSE;
2672    
2673      /* 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
2674      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
2675      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
2676    
2677      class_charcount = 0;      class_charcount = 0;
2678      class_lastchar = -1;      class_lastchar = -1;
2679    
2680        /* Initialize the 32-char bit map to all zeros. We build the map in a
2681        temporary bit of memory, in case the class contains only 1 character (less
2682        than 256), because in that case the compiled code doesn't use the bit map.
2683        */
2684    
2685        memset(classbits, 0, 32 * sizeof(uschar));
2686    
2687  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2688      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2689      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2690        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2691  #endif  #endif
2692    
     /* 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));  
   
2693      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
2694      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
2695      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. */  
2696    
2697      do      if (c != 0) do
2698        {        {
2699          const uschar *oldptr;
2700    
2701  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2702        if (utf8 && c > 127)        if (utf8 && c > 127)
2703          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2704          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2705          }          }
2706    
2707          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2708          data and reset the pointer. This is so that very large classes that
2709          contain a zillion UTF-8 characters no longer overwrite the work space
2710          (which is on the stack). */
2711    
2712          if (lengthptr != NULL)
2713            {
2714            *lengthptr += class_utf8data - class_utf8data_base;
2715            class_utf8data = class_utf8data_base;
2716            }
2717    
2718  #endif  #endif
2719    
2720        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
2721    
2722        if (inescq)        if (inescq)
2723          {          {
2724          if (c == '\\' && ptr[1] == 'E')          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */
2725            {            {
2726            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
2727            ptr++;            ptr++;                            /* Skip the 'E' */
2728            continue;            continue;                         /* Carry on with next */
2729            }            }
2730          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
2731          }          }
2732    
2733        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1876  for (;; ptr++) Line 2738  for (;; ptr++)
2738    
2739        if (c == '[' &&        if (c == '[' &&
2740            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2741            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr))
2742          {          {
2743          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2744          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
# Line 1893  for (;; ptr++) Line 2755  for (;; ptr++)
2755          if (*ptr == '^')          if (*ptr == '^')
2756            {            {
2757            local_negate = TRUE;            local_negate = TRUE;
2758              should_flip_negation = TRUE;  /* Note negative special */
2759            ptr++;            ptr++;
2760            }            }
2761    
# Line 1956  for (;; ptr++) Line 2819  for (;; ptr++)
2819          }          }
2820    
2821        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
2822        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
2823        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.
2824        Inside a class (and only there) it is treated as backspace. Elsewhere        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2825        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  
2826        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2827    
2828        if (c == '\\')        if (c == '\\')
2829          {          {
2830          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2831            if (*errorcodeptr != 0) goto FAILED;
2832    
2833          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2834          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 */
2835            else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2836          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2837            {            {
2838            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == '\\' && ptr[2] == 'E')
# Line 1978  for (;; ptr++) Line 2842  for (;; ptr++)
2842            else inescq = TRUE;            else inescq = TRUE;
2843            continue;            continue;
2844            }            }
2845            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2846    
2847          if (c < 0)          if (c < 0)
2848            {            {
2849            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
2850            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
2851            switch (-c)  
2852              /* Save time by not doing this in the pre-compile phase. */
2853    
2854              if (lengthptr == NULL) switch (-c)
2855              {              {
2856              case ESC_d:              case ESC_d:
2857              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
2858              continue;              continue;
2859    
2860              case ESC_D:              case ESC_D:
2861                should_flip_negation = TRUE;
2862              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2863              continue;              continue;
2864    
# Line 1998  for (;; ptr++) Line 2867  for (;; ptr++)
2867              continue;              continue;
2868    
2869              case ESC_W:              case ESC_W:
2870                should_flip_negation = TRUE;
2871              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2872              continue;              continue;
2873    
# Line 2007  for (;; ptr++) Line 2877  for (;; ptr++)
2877              continue;              continue;
2878    
2879              case ESC_S:              case ESC_S:
2880                should_flip_negation = TRUE;
2881              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2882              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2883              continue;              continue;
2884    
2885  #ifdef SUPPORT_UCP              default:    /* Not recognized; fall through */
2886              case ESC_p:              break;      /* Need "default" setting to stop compiler warning. */
2887              case ESC_P:              }
2888    
2889              /* In the pre-compile phase, just do the recognition. */
2890    
2891              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2892                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
2893    
2894              /* We need to deal with \H, \h, \V, and \v in both phases because
2895              they use extra memory. */
2896    
2897              if (-c == ESC_h)
2898                {
2899                SETBIT(classbits, 0x09); /* VT */
2900                SETBIT(classbits, 0x20); /* SPACE */
2901                SETBIT(classbits, 0xa0); /* NSBP */
2902    #ifdef SUPPORT_UTF8
2903                if (utf8)
2904                {                {
               BOOL negated;  
               int pdata;  
               int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);  
               if (ptype < 0) goto FAILED;  
2905                class_utf8 = TRUE;                class_utf8 = TRUE;
2906                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = XCL_SINGLE;
2907                  XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2908                *class_utf8data++ = ptype;                *class_utf8data++ = XCL_SINGLE;
2909                *class_utf8data++ = pdata;                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2910                class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = XCL_RANGE;
2911                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2912                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2913                  *class_utf8data++ = XCL_SINGLE;
2914                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2915                  *class_utf8data++ = XCL_SINGLE;
2916                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2917                  *class_utf8data++ = XCL_SINGLE;
2918                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2919                }                }
             continue;  
2920  #endif  #endif
2921                continue;
2922                }
2923    
2924              /* Unrecognized escapes are faulted if PCRE is running in its            if (-c == ESC_H)
2925              strict mode. By default, for compatibility with Perl, they are              {
2926              treated as literals. */              for (c = 0; c < 32; c++)
2927                  {
2928                  int x = 0xff;
2929                  switch (c)
2930                    {
2931                    case 0x09/8: x ^= 1 << (0x09%8); break;
2932                    case 0x20/8: x ^= 1 << (0x20%8); break;
2933                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2934                    default: break;
2935                    }
2936                  classbits[c] |= x;
2937                  }
2938    
2939              default:  #ifdef SUPPORT_UTF8
2940              if ((options & PCRE_EXTRA) != 0)              if (utf8)
2941                {                {
2942                *errorcodeptr = ERR7;                class_utf8 = TRUE;
2943                goto FAILED;                *class_utf8data++ = XCL_RANGE;
2944                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2945                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
2946                  *class_utf8data++ = XCL_RANGE;
2947                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
2948                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
2949                  *class_utf8data++ = XCL_RANGE;
2950                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
2951                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
2952                  *class_utf8data++ = XCL_RANGE;
2953                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
2954                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
2955                  *class_utf8data++ = XCL_RANGE;
2956                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
2957                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
2958                  *class_utf8data++ = XCL_RANGE;
2959                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
2960                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
2961                  *class_utf8data++ = XCL_RANGE;
2962                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
2963                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2964                }                }
2965              c = *ptr;              /* The final character */  #endif
2966              class_charcount -= 2;  /* Undo the default count from above */              continue;
2967              }              }
           }  
2968    
2969          /* Fall through if we have a single character (c >= 0). This may be            if (-c == ESC_v)
2970          > 256 in UTF-8 mode. */              {
2971                SETBIT(classbits, 0x0a); /* LF */
2972                SETBIT(classbits, 0x0b); /* VT */
2973                SETBIT(classbits, 0x0c); /* FF */
2974                SETBIT(classbits, 0x0d); /* CR */
2975                SETBIT(classbits, 0x85); /* NEL */
2976    #ifdef SUPPORT_UTF8
2977                if (utf8)
2978                  {
2979                  class_utf8 = TRUE;
2980                  *class_utf8data++ = XCL_RANGE;
2981                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
2982                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
2983                  }
2984    #endif
2985                continue;
2986                }
2987    
2988              if (-c == ESC_V)
2989                {
2990                for (c = 0; c < 32; c++)
2991                  {
2992                  int x = 0xff;
2993                  switch (c)
2994                    {
2995                    case 0x0a/8: x ^= 1 << (0x0a%8);
2996                                 x ^= 1 << (0x0b%8);
2997                                 x ^= 1 << (0x0c%8);
2998                                 x ^= 1 << (0x0d%8);
2999                                 break;
3000                    case 0x85/8: x ^= 1 << (0x85%8); break;
3001                    default: break;
3002                    }
3003                  classbits[c] |= x;
3004                  }
3005    
3006    #ifdef SUPPORT_UTF8
3007                if (utf8)
3008                  {
3009                  class_utf8 = TRUE;
3010                  *class_utf8data++ = XCL_RANGE;
3011                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3012                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3013                  *class_utf8data++ = XCL_RANGE;
3014                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3015                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3016                  }
3017    #endif
3018                continue;
3019                }
3020    
3021              /* We need to deal with \P and \p in both phases. */
3022    
3023    #ifdef SUPPORT_UCP
3024              if (-c == ESC_p || -c == ESC_P)
3025                {
3026                BOOL negated;
3027                int pdata;
3028                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3029                if (ptype < 0) goto FAILED;
3030                class_utf8 = TRUE;
3031                *class_utf8data++ = ((-c == ESC_p) != negated)?
3032                  XCL_PROP : XCL_NOTPROP;
3033                *class_utf8data++ = ptype;
3034                *class_utf8data++ = pdata;
3035                class_charcount -= 2;   /* Not a < 256 character */
3036                continue;
3037                }
3038    #endif
3039              /* Unrecognized escapes are faulted if PCRE is running in its
3040              strict mode. By default, for compatibility with Perl, they are
3041              treated as literals. */
3042    
3043              if ((options & PCRE_EXTRA) != 0)
3044                {
3045                *errorcodeptr = ERR7;
3046                goto FAILED;
3047                }
3048    
3049              class_charcount -= 2;  /* Undo the default count from above */
3050              c = *ptr;              /* Get the final character and fall through */
3051              }
3052    
3053            /* Fall through if we have a single character (c >= 0). This may be
3054            greater than 256 in UTF-8 mode. */
3055    
3056          }   /* End of backslash handling */          }   /* End of backslash handling */
3057    
3058        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3059        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
3060        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3061          entirely. The code for handling \Q and \E is messy. */
3062    
3063          CHECK_RANGE:
3064          while (ptr[1] == '\\' && ptr[2] == 'E')
3065            {
3066            inescq = FALSE;
3067            ptr += 2;
3068            }
3069    
3070          oldptr = ptr;
3071    
3072          /* Remember \r or \n */
3073    
3074        if (ptr[1] == '-' && ptr[2] != ']')        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3075    
3076          /* Check for range */
3077    
3078          if (!inescq && ptr[1] == '-')
3079          {          {
3080          int d;          int d;
3081          ptr += 2;          ptr += 2;
3082            while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;
3083    
3084            /* If we hit \Q (not followed by \E) at this point, go into escaped
3085            mode. */
3086    
3087            while (*ptr == '\\' && ptr[1] == 'Q')
3088              {
3089              ptr += 2;
3090              if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }
3091              inescq = TRUE;
3092              break;
3093              }
3094    
3095            if (*ptr == 0 || (!inescq && *ptr == ']'))
3096              {
3097              ptr = oldptr;
3098              goto LONE_SINGLE_CHARACTER;
3099              }
3100    
3101  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3102          if (utf8)          if (utf8)
# Line 2071  for (;; ptr++) Line 3111  for (;; ptr++)
3111          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
3112          in such circumstances. */          in such circumstances. */
3113    
3114          if (d == '\\')          if (!inescq && d == '\\')
3115            {            {
3116            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3117            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3118    
3119            /* \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
3120            was literal */            special means the '-' was literal */
3121    
3122            if (d < 0)            if (d < 0)
3123              {              {
3124              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = '\b';
3125              else if (d == -ESC_X) d = 'X'; else              else if (d == -ESC_X) d = 'X';
3126                else if (d == -ESC_R) d = 'R'; else
3127                {                {
3128                ptr = oldptr - 2;                ptr = oldptr;
3129                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3130                }                }
3131              }              }
3132            }            }
3133    
3134          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3135          the pre-pass. Optimize one-character ranges */          one-character ranges */
3136    
3137            if (d < c)
3138              {
3139              *errorcodeptr = ERR8;
3140              goto FAILED;
3141              }
3142    
3143          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3144    
3145            /* Remember \r or \n */
3146    
3147            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3148    
3149          /* 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
3150          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3151          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2112  for (;; ptr++) Line 3163  for (;; ptr++)
3163  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3164            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3165              {              {
3166              int occ, ocd;              unsigned int occ, ocd;
3167              int cc = c;              unsigned int cc = c;
3168              int origd = d;              unsigned int origd = d;
3169              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3170                {                {
3171                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3172                      ocd <= (unsigned int)d)
3173                    continue;                          /* Skip embedded ranges */
3174    
3175                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3176                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3177                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3178                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3179                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3180                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3181                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3182                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3183                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3184                  d = ocd;                  d = ocd;
3185                  continue;                  continue;
# Line 2172  for (;; ptr++) Line 3227  for (;; ptr++)
3227          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
3228          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3229    
3230          for (; c <= d; c++)          class_charcount += d - c + 1;
3231            class_lastchar = d;
3232    
3233            /* We can save a bit of time by skipping this in the pre-compile. */
3234    
3235            if (lengthptr == NULL) for (; c <= d; c++)
3236            {            {
3237            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3238            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2180  for (;; ptr++) Line 3240  for (;; ptr++)
3240              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3241              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3242              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3243            }            }
3244    
3245          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2205  for (;; ptr++) Line 3263  for (;; ptr++)
3263  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3264          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3265            {            {
3266            int othercase;            unsigned int othercase;
3267            if ((othercase = _pcre_ucp_othercase(c)) >= 0)            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)
3268              {              {
3269              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3270              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2231  for (;; ptr++) Line 3289  for (;; ptr++)
3289          }          }
3290        }        }
3291    
3292      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3293      loop. This "while" is the end of the "do" above. */  
3294        while ((c = *(++ptr)) != 0 && (c != ']' || inescq));
3295    
3296        if (c == 0)                          /* Missing terminating ']' */
3297          {
3298          *errorcodeptr = ERR6;
3299          goto FAILED;
3300          }
3301    
3302    
3303    /* This code has been disabled because it would mean that \s counts as
3304    an explicit \r or \n reference, and that's not really what is wanted. Now
3305    we set the flag only if there is a literal "\r" or "\n" in the class. */
3306    
3307    #if 0
3308        /* Remember whether \r or \n are in this class */
3309    
3310        if (negate_class)
3311          {
3312          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3313          }
3314        else
3315          {
3316          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3317          }
3318    #endif
3319    
     while ((c = *(++ptr)) != ']' || inescq);  
3320    
3321      /* 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
3322      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
3323      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
3324      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3325      single-bytes only. This is an historical hangover. Maybe one day we can  
3326      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3327        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3328        operate on single-bytes only. This is an historical hangover. Maybe one day
3329        we can tidy these opcodes to handle multi-byte characters.
3330    
3331      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
3332      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 2251  for (;; ptr++) Line 3336  for (;; ptr++)
3336      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3337    
3338  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3339      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3340            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3341  #else  #else
3342      if (class_charcount == 1)      if (class_charcount == 1)
3343  #endif  #endif
# Line 2297  for (;; ptr++) Line 3380  for (;; ptr++)
3380      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3381    
3382      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3383      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3384      we can omit the bitmap. */      such as \S in the class, because in that case all characters > 255 are in
3385        the class, so any that were explicitly given as well can be ignored. If
3386        (when there are explicit characters > 255 that must be listed) there are no
3387        characters < 256, we can omit the bitmap in the actual compiled code. */
3388    
3389  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3390      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3391        {        {
3392        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3393        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3394        code += LINK_SIZE;        code += LINK_SIZE;
3395        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3396    
3397        /* 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;
3398        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3399    
3400        if (class_charcount > 0)        if (class_charcount > 0)
3401          {          {
3402          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3403            memmove(code + 32, code, class_utf8data - code);
3404          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3405          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;  
3406          }          }
3407          else code = class_utf8data;
3408    
3409        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3410    
# Line 2334  for (;; ptr++) Line 3413  for (;; ptr++)
3413        }        }
3414  #endif  #endif
3415    
3416      /* 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
3417      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
3418      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
3419      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3420    
3421        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3422      if (negate_class)      if (negate_class)
3423        {        {
3424        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3425        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3426        }        }
3427      else      else
3428        {        {
       *code++ = OP_CLASS;  
3429        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3430        }        }
3431      code += 32;      code += 32;
3432      break;      break;
3433    
3434    
3435        /* ===================================================================*/
3436      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3437      has been tested above. */      has been tested above. */
3438    
# Line 2419  for (;; ptr++) Line 3500  for (;; ptr++)
3500        }        }
3501      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3502    
     /* 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;  
       }  
   
3503      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3504      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
3505      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 2466  for (;; ptr++) Line 3533  for (;; ptr++)
3533          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3534          }          }
3535    
3536          /* If the repetition is unlimited, it pays to see if the next thing on
3537          the line is something that cannot possibly match this character. If so,
3538          automatically possessifying this item gains some performance in the case
3539          where the match fails. */
3540    
3541          if (!possessive_quantifier &&
3542              repeat_max < 0 &&
3543              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
3544                options, cd))
3545            {
3546            repeat_type = 0;    /* Force greedy */
3547            possessive_quantifier = TRUE;
3548            }
3549    
3550        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
3551        }        }
3552    
3553      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
3554      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-
3555      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
3556      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
3557        currently used only for single-byte chars. */
3558    
3559      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
3560        {        {
3561        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
3562        c = previous[1];        c = previous[1];
3563          if (!possessive_quantifier &&
3564              repeat_max < 0 &&
3565              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
3566            {
3567            repeat_type = 0;    /* Force greedy */
3568            possessive_quantifier = TRUE;
3569            }
3570        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
3571        }        }
3572    
# Line 2495  for (;; ptr++) Line 3584  for (;; ptr++)
3584        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
3585        c = *previous;        c = *previous;
3586    
3587          if (!possessive_quantifier &&
3588              repeat_max < 0 &&
3589              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
3590            {
3591            repeat_type = 0;    /* Force greedy */
3592            possessive_quantifier = TRUE;
3593            }
3594    
3595        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
3596        if (*previous == OP_PROP || *previous == OP_NOTPROP)        if (*previous == OP_PROP || *previous == OP_NOTPROP)
3597          {          {
# Line 2514  for (;; ptr++) Line 3611  for (;; ptr++)
3611        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3612        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3613    
3614        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3615    
3616        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3617    
# Line 2535  for (;; ptr++) Line 3632  for (;; ptr++)
3632          }          }
3633    
3634        /* 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
3635        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
3636        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
3637        one less than the maximum. */        one less than the maximum. */
3638    
# Line 2588  for (;; ptr++) Line 3685  for (;; ptr++)
3685            }            }
3686    
3687          /* 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
3688          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
3689            UPTO is just for 1 instance, we can use QUERY instead. */
3690    
3691          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
3692            {            {
# Line 2607  for (;; ptr++) Line 3705  for (;; ptr++)
3705              *code++ = prop_value;              *code++ = prop_value;
3706              }              }
3707            repeat_max -= repeat_min;            repeat_max -= repeat_min;
3708            *code++ = OP_UPTO + repeat_type;  
3709            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
3710                {
3711                *code++ = OP_QUERY + repeat_type;
3712                }
3713              else
3714                {
3715                *code++ = OP_UPTO + repeat_type;
3716                PUT2INC(code, 0, repeat_max);
3717                }
3718            }            }
3719          }          }
3720    
# Line 2655  for (;; ptr++) Line 3761  for (;; ptr++)
3761        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3762        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3763    
3764        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3765    
3766        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3767          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2675  for (;; ptr++) Line 3781  for (;; ptr++)
3781      /* 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
3782      cases. */      cases. */
3783    
3784      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
3785               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
3786        {        {
3787        register int i;        register int i;
3788        int ketoffset = 0;        int ketoffset = 0;
3789        int len = code - previous;        int len = code - previous;
3790        uschar *bralink = NULL;        uschar *bralink = NULL;
3791    
3792          /* Repeating a DEFINE group is pointless */
3793    
3794          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
3795            {
3796            *errorcodeptr = ERR55;
3797            goto FAILED;
3798            }
3799    
3800        /* 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
3801        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
3802        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 2717  for (;; ptr++) Line 3831  for (;; ptr++)
3831          /* If the maximum is 1 or unlimited, we just have to stick in the          /* If the maximum is 1 or unlimited, we just have to stick in the
3832          BRAZERO and do no more at this point. However, we do need to adjust          BRAZERO and do no more at this point. However, we do need to adjust
3833          any OP_RECURSE calls inside the group that refer to the group itself or          any OP_RECURSE calls inside the group that refer to the group itself or
3834          any internal group, because the offset is from the start of the whole          any internal or forward referenced group, because the offset is from
3835          regex. Temporarily terminate the pattern while doing this. */          the start of the whole regex. Temporarily terminate the pattern while
3836            doing this. */
3837    
3838          if (repeat_max <= 1)          if (repeat_max <= 1)
3839            {            {
3840            *code = OP_END;            *code = OP_END;
3841            adjust_recurse(previous, 1, utf8, cd);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3842            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3843            code++;            code++;
3844            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2741  for (;; ptr++) Line 3856  for (;; ptr++)
3856            {            {
3857            int offset;            int offset;
3858            *code = OP_END;            *code = OP_END;
3859            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd);            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
3860            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, len);
3861            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
3862            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 2761  for (;; ptr++) Line 3876  for (;; ptr++)
3876        /* If the minimum is greater than zero, replicate the group as many        /* If the minimum is greater than zero, replicate the group as many
3877        times as necessary, and adjust the maximum to the number of subsequent        times as necessary, and adjust the maximum to the number of subsequent
3878        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
3879        set a required char, copy the latter from the former. */        set a required char, copy the latter from the former. If there are any
3880          forward reference subroutine calls in the group, there will be entries on
3881          the workspace list; replicate these with an appropriate increment. */
3882    
3883        else        else
3884          {          {
3885          if (repeat_min > 1)          if (repeat_min > 1)
3886            {            {
3887            if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;            /* In the pre-compile phase, we don't actually do the replication. We
3888            for (i = 1; i < repeat_min; i++)            just adjust the length as if we had. Do some paranoid checks for
3889              potential integer overflow. */
3890    
3891              if (lengthptr != NULL)
3892                {
3893                int delta = (repeat_min - 1)*length_prevgroup;
3894                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3895                                                                (double)INT_MAX ||
3896                    OFLOW_MAX - *lengthptr < delta)
3897                  {
3898                  *errorcodeptr = ERR20;
3899                  goto FAILED;
3900                  }
3901                *lengthptr += delta;
3902                }
3903    
3904              /* This is compiling for real */
3905    
3906              else
3907              {              {
3908              memcpy(code, previous, len);              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
3909              code += len;              for (i = 1; i < repeat_min; i++)
3910                  {
3911                  uschar *hc;
3912                  uschar *this_hwm = cd->hwm;
3913                  memcpy(code, previous, len);
3914                  for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3915                    {
3916                    PUT(cd->hwm, 0, GET(hc, 0) + len);
3917                    cd->hwm += LINK_SIZE;
3918                    }
3919                  save_hwm = this_hwm;
3920                  code += len;
3921                  }
3922              }              }
3923            }            }
3924    
3925          if (repeat_max > 0) repeat_max -= repeat_min;          if (repeat_max > 0) repeat_max -= repeat_min;
3926          }          }
3927    
# Line 2781  for (;; ptr++) Line 3929  for (;; ptr++)
3929        the maximum is limited, it replicates the group in a nested fashion,        the maximum is limited, it replicates the group in a nested fashion,
3930        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,
3931        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
3932        the number of additional copies needed. */        the number of additional copies needed. Again, we must remember to
3933          replicate entries on the forward reference list. */
3934    
3935        if (repeat_max >= 0)        if (repeat_max >= 0)
3936          {          {
3937          for (i = repeat_max - 1; i >= 0; i--)          /* In the pre-compile phase, we don't actually do the replication. We
3938            just adjust the length as if we had. For each repetition we must add 1
3939            to the length for BRAZERO and for all but the last repetition we must
3940            add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3941            paranoid checks to avoid integer overflow. */
3942    
3943            if (lengthptr != NULL && repeat_max > 0)
3944              {
3945              int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3946                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3947              if ((double)repeat_max *
3948                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3949                      > (double)INT_MAX ||
3950                  OFLOW_MAX - *lengthptr < delta)
3951                {
3952                *errorcodeptr = ERR20;
3953                goto FAILED;
3954                }
3955              *lengthptr += delta;
3956              }
3957    
3958            /* This is compiling for real */
3959    
3960            else for (i = repeat_max - 1; i >= 0; i--)
3961            {            {
3962              uschar *hc;
3963              uschar *this_hwm = cd->hwm;
3964    
3965            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
3966    
3967            /* All but the final copy start a new nesting, maintaining the            /* All but the final copy start a new nesting, maintaining the
# Line 2802  for (;; ptr++) Line 3977  for (;; ptr++)
3977              }              }
3978    
3979            memcpy(code, previous, len);            memcpy(code, previous, len);
3980              for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
3981                {
3982                PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
3983                cd->hwm += LINK_SIZE;
3984                }
3985              save_hwm = this_hwm;
3986            code += len;            code += len;
3987            }            }
3988    
# Line 2824  for (;; ptr++) Line 4005  for (;; ptr++)
4005        /* 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
4006        can't just offset backwards from the current code point, because we        can't just offset backwards from the current code point, because we
4007        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
4008        correct offset was computed above. */        correct offset was computed above.
4009    
4010          Then, when we are doing the actual compile phase, check to see whether
4011          this group is a non-atomic one that could match an empty string. If so,
4012          convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
4013          that runtime checking can be done. [This check is also applied to
4014          atomic groups at runtime, but in a different way.] */
4015    
4016        else code[-ketoffset] = OP_KETRMAX + repeat_type;        else
4017            {
4018            uschar *ketcode = code - ketoffset;
4019            uschar *bracode = ketcode - GET(ketcode, 1);
4020            *ketcode = OP_KETRMAX + repeat_type;
4021            if (lengthptr == NULL && *bracode != OP_ONCE)
4022              {
4023              uschar *scode = bracode;
4024              do
4025                {
4026                if (could_be_empty_branch(scode, ketcode, utf8))
4027                  {
4028                  *bracode += OP_SBRA - OP_BRA;
4029                  break;
4030                  }
4031                scode += GET(scode, 1);
4032                }
4033              while (*scode == OP_ALT);
4034              }
4035            }
4036        }        }
4037    
4038      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 2837  for (;; ptr++) Line 4043  for (;; ptr++)
4043        goto FAILED;        goto FAILED;
4044        }        }
4045    
4046      /* If the character following a repeat is '+', we wrap the entire repeated      /* If the character following a repeat is '+', or if certain optimization
4047      item inside OP_ONCE brackets. This is just syntactic sugar, taken from      tests above succeeded, possessive_quantifier is TRUE. For some of the
4048      Sun's Java package. The repeated item starts at tempcode, not at previous,      simpler opcodes, there is an special alternative opcode for this. For
4049      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.
4050      repeated. However, we don't support '+' after a greediness '?'. */      The '+' notation is just syntactic sugar, taken from Sun's Java package,
4051        but the special opcodes can optimize it a bit. The repeated item starts at
4052        tempcode, not at previous, which might be the first part of a string whose
4053        (former) last char we repeated.
4054    
4055        Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4056        an 'upto' may follow. We skip over an 'exact' item, and then test the
4057        length of what remains before proceeding. */
4058    
4059      if (possessive_quantifier)      if (possessive_quantifier)
4060        {        {
4061        int len = code - tempcode;        int len;
4062        memmove(tempcode + 1+LINK_SIZE, tempcode, len);        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4063        code += 1 + LINK_SIZE;            *tempcode == OP_NOTEXACT)
4064        len += 1 + LINK_SIZE;          tempcode += _pcre_OP_lengths[*tempcode] +
4065        tempcode[0] = OP_ONCE;            ((*tempcode == OP_TYPEEXACT &&
4066        *code++ = OP_KET;               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4067        PUTINC(code, 0, len);        len = code - tempcode;
4068        PUT(tempcode, 1, len);        if (len > 0) switch (*tempcode)
4069            {
4070            case OP_STAR:  *tempcode = OP_POSSTAR; break;
4071            case OP_PLUS:  *tempcode = OP_POSPLUS; break;
4072            case OP_QUERY: *tempcode = OP_POSQUERY; break;
4073            case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4074    
4075            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4076            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4077            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4078            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4079    
4080            case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4081            case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4082            case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4083            case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4084    
4085            default:
4086            memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4087            code += 1 + LINK_SIZE;
4088            len += 1 + LINK_SIZE;
4089            tempcode[0] = OP_ONCE;
4090            *code++ = OP_KET;
4091            PUTINC(code, 0, len);
4092            PUT(tempcode, 1, len);
4093            break;
4094            }
4095        }        }
4096    
4097      /* 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 2865  for (;; ptr++) Line 4104  for (;; ptr++)
4104      break;      break;
4105    
4106    
4107      /* Start of nested bracket sub-expression, or comment or lookahead or      /* ===================================================================*/
4108      lookbehind or option setting or condition. First deal with special things      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4109      that can come after a bracket; all are introduced by ?, and the appearance      lookbehind or option setting or condition or all the other extended
4110      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.  */  
4111    
4112      case '(':      case '(':
4113      newoptions = options;      newoptions = options;
4114      skipbytes = 0;      skipbytes = 0;
4115        bravalue = OP_CBRA;
4116        save_hwm = cd->hwm;
4117        reset_bracount = FALSE;
4118    
4119        /* First deal with various "verbs" that can be introduced by '*'. */
4120    
4121        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4122          {
4123          int i, namelen;
4124          const char *vn = verbnames;
4125          const uschar *name = ++ptr;
4126          previous = NULL;
4127          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4128          if (*ptr == ':')
4129            {
4130            *errorcodeptr = ERR59;   /* Not supported */
4131            goto FAILED;
4132            }
4133          if (*ptr != ')')
4134            {
4135            *errorcodeptr = ERR60;
4136            goto FAILED;
4137            }
4138          namelen = ptr - name;
4139          for (i = 0; i < verbcount; i++)
4140            {
4141            if (namelen == verbs[i].len &&
4142                strncmp((char *)name, vn, namelen) == 0)
4143              {
4144              *code = verbs[i].op;
4145              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4146              break;
4147              }
4148            vn += verbs[i].len + 1;
4149            }
4150          if (i < verbcount) continue;
4151          *errorcodeptr = ERR60;
4152          goto FAILED;
4153          }
4154    
4155      if (*(++ptr) == '?')      /* Deal with the extended parentheses; all are introduced by '?', and the
4156        appearance of any of them means that this is not a capturing group. */
4157    
4158        else if (*ptr == '?')
4159        {        {
4160        int set, unset;        int i, set, unset, namelen;
4161        int *optset;        int *optset;
4162          const uschar *name;
4163          uschar *slot;
4164    
4165        switch (*(++ptr))        switch (*(++ptr))
4166          {          {
4167          case '#':                 /* Comment; skip to ket */          case '#':                 /* Comment; skip to ket */
4168          ptr++;          ptr++;
4169          while (*ptr != ')') ptr++;          while (*ptr != 0 && *ptr != ')') ptr++;
4170            if (*ptr == 0)
4171              {
4172              *errorcodeptr = ERR18;
4173              goto FAILED;
4174              }
4175          continue;          continue;
4176    
4177          case ':':                 /* Non-extracting bracket */  
4178            /* ------------------------------------------------------------ */
4179            case '|':                 /* Reset capture count for each branch */
4180            reset_bracount = TRUE;
4181            /* Fall through */
4182    
4183            /* ------------------------------------------------------------ */
4184            case ':':                 /* Non-capturing bracket */
4185          bravalue = OP_BRA;          bravalue = OP_BRA;
4186          ptr++;          ptr++;
4187          break;          break;
4188    
4189    
4190            /* ------------------------------------------------------------ */
4191          case '(':          case '(':
4192          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4193    
4194          /* A condition can be a number, referring to a numbered group, a name,          /* A condition can be an assertion, a number (referring to a numbered
4195          referring to a named group, 'R', referring to recursion, or an          group), a name (referring to a named group), or 'R', referring to
4196          assertion. There are two unfortunate ambiguities, caused by history.          recursion. R<digits> and R&name are also permitted for recursion tests.
4197          (a) 'R' can be the recursive thing or the name 'R', and (b) a number  
4198          could be a name that consists of digits. In both cases, we look for a          There are several syntaxes for testing a named group: (?(name)) is used
4199          name first; if not found, we try the other cases. If the first          by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
4200          character after (?( is a word character, we know the rest up to ) will  
4201          also be word characters because the syntax was checked in the first          There are two unfortunate ambiguities, caused by history. (a) 'R' can
4202          pass. */          be the recursive thing or the name 'R' (and similarly for 'R' followed
4203            by digits), and (b) a number could be a name that consists of digits.
4204          if ((cd->ctypes[ptr[1]] & ctype_word) != 0)          In both cases, we look for a name first; if not found, we try the other
4205            {          cases. */
4206            int i, namelen;  
4207            int condref = 0;          /* For conditions that are assertions, check the syntax, and then exit
4208            const uschar *name;          the switch. This will take control down to where bracketed groups,
4209            uschar *slot = cd->name_table;          including assertions, are processed. */
4210    
4211            /* This is needed for all successful cases. */          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))
4212              break;
4213    
4214            skipbytes = 3;          /* Most other conditions use OP_CREF (a couple change to OP_RREF
4215            below), and all need to skip 3 bytes at the start of the group. */
4216    
4217            /* Read the name, but also get it as a number if it's all digits */          code[1+LINK_SIZE] = OP_CREF;
4218            skipbytes = 3;
4219            refsign = -1;
4220    
4221            name = ++ptr;          /* Check for a test for recursion in a named group. */
4222            while (*ptr != ')')  
4223              {          if (ptr[1] == 'R' && ptr[2] == '&')
4224              if (condref >= 0)            {
4225                condref = ((digitab[*ptr] & ctype_digit) != 0)?            terminator = -1;
4226                  condref * 10 + *ptr - '0' : -1;            ptr += 2;
4227              ptr++;            code[1+LINK_SIZE] = OP_RREF;    /* Change the type of test */
4228              }            }
4229            namelen = ptr - name;  
4230            /* Check for a test for a named group's having been set, using the Perl
4231            syntax (?(<name>) or (?('name') */
4232    
4233            else if (ptr[1] == '<')
4234              {
4235              terminator = '>';
4236              ptr++;
4237              }
4238            else if (ptr[1] == '\'')
4239              {
4240              terminator = '\'';
4241            ptr++;            ptr++;
4242              }
4243            else
4244              {
4245              terminator = 0;
4246              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4247              }
4248    
4249            for (i = 0; i < cd->names_found; i++)          /* We now expect to read a name; any thing else is an error */
             {  
             if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;  
             slot += cd->name_entry_size;  
             }  
4250    
4251            /* Found a previous named subpattern */          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
4252              {
4253              ptr += 1;  /* To get the right offset */
4254              *errorcodeptr = ERR28;
4255              goto FAILED;
4256              }
4257    
4258            if (i < cd->names_found)          /* Read the name, but also get it as a number if it's all digits */
             {  
             condref = GET2(slot, 0);  
             code[1+LINK_SIZE] = OP_CREF;  
             PUT2(code, 2+LINK_SIZE, condref);  
             }  
4259    
4260            /* Search the pattern for a forward reference */          recno = 0;
4261            name = ++ptr;
4262            while ((cd->ctypes[*ptr] & ctype_word) != 0)
4263              {
4264              if (recno >= 0)
4265                recno = ((digitab[*ptr] & ctype_digit) != 0)?
4266                  recno * 10 + *ptr - '0' : -1;
4267              ptr++;
4268              }
4269            namelen = ptr - name;
4270    
4271            else if ((i = find_named_parens(ptr, *brackets, name, namelen)) > 0)          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')
4272              {            {
4273              code[1+LINK_SIZE] = OP_CREF;            ptr--;      /* Error offset */
4274              PUT2(code, 2+LINK_SIZE, i);            *errorcodeptr = ERR26;
4275              }            goto FAILED;
4276              }
4277    
4278            /* Check for 'R' for recursion */          /* Do no further checking in the pre-compile phase. */
4279    
4280            else if (namelen == 1 && *name == 'R')          if (lengthptr != NULL) break;
             {  
             code[1+LINK_SIZE] = OP_CREF;  
             PUT2(code, 2+LINK_SIZE, CREF_RECURSE);  
             }  
4281    
4282            /* Check for a subpattern number */          /* In the real compile we do the work of looking for the actual
4283            reference. If the string started with "+" or "-" we require the rest to
4284            be digits, in which case recno will be set. */
4285    
4286            else if (condref > 0)          if (refsign > 0)
4287              {
4288              if (recno <= 0)
4289              {              {
4290              code[1+LINK_SIZE] = OP_CREF;              *errorcodeptr = ERR58;
4291              PUT2(code, 2+LINK_SIZE, condref);              goto FAILED;
4292              }              }
4293              recno = (refsign == '-')?
4294            /* Either an unidentified subpattern, or a reference to (?(0) */              cd->bracount - recno + 1 : recno +cd->bracount;
4295              if (recno <= 0 || recno > cd->final_bracount)
           else  
4296              {              {
4297              *errorcodeptr = (condref == 0)? ERR35: ERR15;              *errorcodeptr = ERR15;
4298              goto FAILED;              goto FAILED;
4299              }              }
4300              PUT2(code, 2+LINK_SIZE, recno);
4301              break;
4302            }            }
4303    
4304          /* For conditions that are assertions, we just fall through, having          /* Otherwise (did not start with "+" or "-"), start by looking for the
4305          set bravalue above. */          name. */
4306    
4307          break;          slot = cd->name_table;
4308            for (i = 0; i < cd->names_found; i++)
4309              {
4310              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
4311              slot += cd->name_entry_size;
4312              }
4313    
4314          case '=':                 /* Positive lookahead */          /* Found a previous named subpattern */
         bravalue = OP_ASSERT;  
         ptr++;  
         break;  
4315    
4316          case '!':                 /* Negative lookahead */          if (i < cd->names_found)
4317          bravalue = OP_ASSERT_NOT;            {
4318          ptr++;            recno = GET2(slot, 0);
4319          break;            PUT2(code, 2+LINK_SIZE, recno);
4320              }
4321    
4322          case '<':                 /* Lookbehinds */          /* Search the pattern for a forward reference */
4323          switch (*(++ptr))  
4324            else if ((i = find_parens(ptr, cd->bracount, name, namelen,
4325                            (options & PCRE_EXTENDED) != 0)) > 0)
4326            {            {
4327            case '=':               /* Positive lookbehind */            PUT2(code, 2+LINK_SIZE, i);
4328            bravalue = OP_ASSERTBACK;            }
           ptr++;  
           break;  
4329    
4330            case '!':               /* Negative lookbehind */          /* If terminator == 0 it means that the name followed directly after
4331            bravalue = OP_ASSERTBACK_NOT;          the opening parenthesis [e.g. (?(abc)...] and in this case there are
4332            ptr++;          some further alternatives to try. For the cases where terminator != 0
4333            break;          [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
4334            now checked all the possibilities, so give an error. */
4335    
4336            else if (terminator != 0)
4337              {
4338              *errorcodeptr = ERR15;
4339              goto FAILED;
4340            }            }
         break;  
4341    
4342          case '>':                 /* One-time brackets */          /* Check for (?(R) for recursion. Allow digits after R to specify a
4343          bravalue = OP_ONCE;          specific group number. */
         ptr++;  
         break;  
4344    
4345          case 'C':                 /* Callout - may be followed by digits; */          else if (*name == 'R')
4346          previous_callout = code;  /* Save for later completion */            {
4347          after_manual_callout = 1; /* Skip one item before completing */            recno = 0;
4348          *code++ = OP_CALLOUT;     /* Already checked that the terminating */            for (i = 1; i < namelen; i++)
4349            {                       /* closing parenthesis is present. */              {
4350            int n = 0;              if ((digitab[name[i]] & ctype_digit) == 0)
4351                  {
4352                  *errorcodeptr = ERR15;
4353                  goto FAILED;
4354                  }
4355                recno = recno * 10 + name[i] - '0';
4356                }
4357              if (recno == 0) recno = RREF_ANY;
4358              code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
4359              PUT2(code, 2+LINK_SIZE, recno);
4360              }
4361    
4362            /* Similarly, check for the (?(DEFINE) "condition", which is always
4363            false. */
4364    
4365            else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)
4366              {
4367              code[1+LINK_SIZE] = OP_DEF;
4368              skipbytes = 1;
4369              }
4370    
4371            /* Check for the "name" actually being a subpattern number. We are
4372            in the second pass here, so final_bracount is set. */
4373    
4374            else if (recno > 0 && recno <= cd->final_bracount)
4375              {
4376              PUT2(code, 2+LINK_SIZE, recno);
4377              }
4378    
4379            /* Either an unidentified subpattern, or a reference to (?(0) */
4380    
4381            else
4382              {
4383              *errorcodeptr = (recno == 0)? ERR35: ERR15;
4384              goto FAILED;
4385              }
4386            break;
4387    
4388    
4389            /* ------------------------------------------------------------ */
4390            case '=':                 /* Positive lookahead */
4391            bravalue = OP_ASSERT;
4392            ptr++;
4393            break;
4394    
4395    
4396            /* ------------------------------------------------------------ */
4397            case '!':                 /* Negative lookahead */
4398            ptr++;
4399            if (*ptr == ')')          /* Optimize (?!) */
4400              {
4401              *code++ = OP_FAIL;
4402              previous = NULL;
4403              continue;
4404              }
4405            bravalue = OP_ASSERT_NOT;
4406            break;
4407    
4408    
4409            /* ------------------------------------------------------------ */
4410            case '<':                 /* Lookbehind or named define */
4411            switch (ptr[1])
4412              {
4413              case '=':               /* Positive lookbehind */
4414              bravalue = OP_ASSERTBACK;
4415              ptr += 2;
4416              break;
4417    
4418              case '!':               /* Negative lookbehind */
4419              bravalue = OP_ASSERTBACK_NOT;
4420              ptr += 2;
4421              break;
4422    
4423              default:                /* Could be name define, else bad */
4424              if ((cd->ctypes[ptr[1]] & ctype_word) != 0) goto DEFINE_NAME;
4425              ptr++;                  /* Correct offset for error */
4426              *errorcodeptr = ERR24;
4427              goto FAILED;
4428              }
4429            break;
4430    
4431    
4432            /* ------------------------------------------------------------ */
4433            case '>':                 /* One-time brackets */
4434            bravalue = OP_ONCE;
4435            ptr++;
4436            break;
4437    
4438    
4439            /* ------------------------------------------------------------ */
4440            case 'C':                 /* Callout - may be followed by digits; */
4441            previous_callout = code;  /* Save for later completion */
4442            after_manual_callout = 1; /* Skip one item before completing */
4443            *code++ = OP_CALLOUT;
4444              {
4445              int n = 0;
4446            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4447              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - '0';
4448              if (*ptr != ')')
4449                {
4450                *errorcodeptr = ERR39;
4451                goto FAILED;
4452                }
4453            if (n > 255)            if (n > 255)
4454              {              {
4455              *errorcodeptr = ERR38;              *errorcodeptr = ERR38;
# Line 3034  for (;; ptr++) Line 4463  for (;; ptr++)
4463          previous = NULL;          previous = NULL;
4464          continue;          continue;
4465    
4466          case 'P':                 /* Named subpattern handling */  
4467          if (*(++ptr) == '<')      /* Definition */          /* ------------------------------------------------------------ */
4468            case 'P':                 /* Python-style named subpattern handling */
4469            if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */
4470              {
4471              is_recurse = *ptr == '>';
4472              terminator = ')';
4473              goto NAMED_REF_OR_RECURSE;
4474              }
4475            else if (*ptr != '<')    /* Test for Python-style definition */
4476            {            {
4477            int i, namelen;            *errorcodeptr = ERR41;
4478            uschar *slot = cd->name_table;            goto FAILED;
4479            const uschar *name;     /* Don't amalgamate; some compilers */            }
4480            name = ++ptr;           /* grumble at autoincrement in declaration */          /* Fall through to handle (?P< as (?< is handled */
4481    
           while (*ptr++ != '>');  
           namelen = ptr - name - 1;  
4482    
4483            for (i = 0; i < cd->names_found; i++)          /* ------------------------------------------------------------ */
4484            DEFINE_NAME:    /* Come here from (?< handling */
4485            case '\'':
4486              {
4487              terminator = (*ptr == '<')? '>' : '\'';
4488              name = ++ptr;
4489    
4490              while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
4491              namelen = ptr - name;
4492    
4493              /* In the pre-compile phase, just do a syntax check. */
4494    
4495              if (lengthptr != NULL)
4496              {              {
4497              int crc = memcmp(name, slot+2, namelen);              if (*ptr != terminator)
4498              if (crc == 0)                {
4499         &nb