Use __LONG_LONG_MAX__ instead of LONG_LONG_MAX in test

	* testsuite/decimal/conversion-to-integral.cc: Use predefined macro
	instead of non-standard glibc one.

From-SVN: r253894

libstdc++-v3/ChangeLog

+2017-10-19  Jonathan Wakely  <jwakely@redhat.com>
+
+	* testsuite/decimal/conversion-to-integral.cc: Use predefined macro
+	instead of non-standard glibc one.
+
 2017-10-16  François Dumont  <fdumont@gcc.gnu.org>
 
 	PR libstdc++/82558

libstdc++-v3/testsuite/decimal/conversion-to-integral.cc

@@ -64,7 +64,7 @@ void
 conversion_to_integral_128 (void)
 {
   #undef MAXVAL
-  #define MAXVAL LONG_LONG_MAX
+  #define MAXVAL __LONG_LONG_MAX__
   decimal128 a, b (1), c (-1), d (MAXVAL), e (-MAXVAL);
   long long ll;