* regex/regcomp.c (wgetnext): Add a kludge to be more glibc compatible.

Add comment to explain.

winsup/cygwin/ChangeLog

@@ -1,3 +1,8 @@
+2013-07-21  Corinna Vinschen  <corinna@vinschen.de>
+
+	* regex/regcomp.c (wgetnext): Add a kludge to be more glibc compatible.
+	Add comment to explain.
+
 2013-07-20  Christopher Faylor  <me.cygwin2013@cgf.cx>
 
 	* exceptions.cc (signal_exit): Remove comment.  Use

winsup/cygwin/regex/regcomp.c

@@ -1128,6 +1128,18 @@ wgetnext(struct parse *p)
 	wint_t ret;
 	size_t n;
 
+#ifdef __CYGWIN__
+	/* Kludge for more glibc compatibility.  On Cygwin as well as on
+	   Linux, mbrtowc returns -1 if the current local's codeset is ASCII
+	   and the character is >= 0x80.  Nevertheless, glibc's regcomp allows
+	   any char value, even stuff like [\xc0-\xff], if the locale's codeset
+	   is ASCII, so in regcomp it ignores the fact that chars >= 0x80 are
+	   invalid ASCII chars.  To be more Linux-compatible, we align the
+	   behaviour to glibc here.  Allow any character value if the current
+	   local's codeset is ASCII. */
+	if (*__locale_charset () == 'A') /* SCII */
+	  return (wint_t) (unsigned char) *p->next++;
+#endif
 	memset(&mbs, 0, sizeof(mbs));
 	n = mbrtowc(&wc, p->next, p->end - p->next, &mbs);
 	if (n == (size_t)-1 || n == (size_t)-2) {

winsup/cygwin/release/1.7.22

@@ -11,6 +11,11 @@ What changed:
 - Added cygwin GetCommandLine wrappers which will allow Cygwin programs to
   (appear to) use the Windows command line functions.
 
+- regcomp(3) now allows character values >= 0x80 if the current codeset is
+  ASCII (default codeset in the "C"/"POSIX" locale).  This allows patterns
+  containing arbitrary byte values as GLibc's regcomp.
+  Fixes: http://cygwin.com/ml/cygwin/2013-07/msg00405.html
+
 
 Bug fixes:
 ----------