Fix [RT#6266] -- sv_pos_u2b expects to be called with a valid character index

[p5sagit/p5-mst-13.2.git] / regcomp.h

diff --git a/regcomp.h b/regcomp.h
index dee7d78..2ac1be1 100644 (file)
--- a/regcomp.h
+++ b/regcomp.h
@@ -18,6 +18,24 @@ typedef OP OP_4tree;                 /* Will be redefined later. */
 /* Be really agressive about optimising patterns with trie sequences? */
 #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1
 
+/* Use old style unicode mappings for perl and posix character classes
+ *
+ * NOTE: Enabling this essentially breaks character class matching against unicode 
+ * strings, so that POSIX char classes match when they shouldn't, and \d matches 
+ * way more than 10 characters, and sometimes a charclass and its complement either
+ * both match or neither match.
+ * NOTE: Disabling this will cause various backwards compatibility issues to rear 
+ * their head, and tests to fail. However it will make the charclass behaviour 
+ * consistant regardless of internal string type, and make character class inversions
+ * consistant. The tests that fail in the regex engine are basically broken tests.
+ *
+ * Personally I think 5.12 should disable this for sure. Its a bit more debatable for
+ * 5.10, so for now im leaving it enabled.
+ *
+ * -demerphq
+ */
+#define PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS 1
+
 /* Should the optimiser take positive assertions into account? */
 #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0
 
@@ -317,9 +335,9 @@ struct regnode_charclass_class {    /* has [[:blah:]] classes */
 #define ANYOF_NALNUM    1
 #define ANYOF_SPACE     2      /* \s */
 #define ANYOF_NSPACE    3
-#define ANYOF_DIGIT     4
+#define ANYOF_DIGIT     4      /* \d */
 #define ANYOF_NDIGIT    5
-#define ANYOF_ALNUMC    6      /* isalnum(3), utf8::IsAlnum, ALNUMC */
+#define ANYOF_ALNUMC    6      /* [[:alnum:]] isalnum(3), utf8::IsAlnum, ALNUMC */
 #define ANYOF_NALNUMC   7
 #define ANYOF_ALPHA     8
 #define ANYOF_NALPHA    9