Follow-up on a149d118.
authorJarkko Hietaniemi <jhi@iki.fi>
Sun, 14 Aug 2016 14:16:55 +0000 (10:16 -0400)
committerJarkko Hietaniemi <jhi@iki.fi>
Sun, 14 Aug 2016 14:16:55 +0000 (10:16 -0400)
The added tests were meant only for x86-80-bit long doubles,
but were accidentally run also on 128-bit doubles (quadmath,
or probably also on true IEEE 754 quadruples).

t/op/hexfp.t

index d6c0482..c4c08cc 100644 (file)
@@ -245,8 +245,11 @@ SKIP:
 
 # [perl #128919] limited exponent range in hex fp literal with long double
 SKIP: {
-    skip("nv_preserves_uv_bits is $Config{nv_preserves_uv_bits} not 64", 4)
-        unless ($Config{nv_preserves_uv_bits} == 64);
+    skip("non-80-bit-long-double", 4)
+        unless ($Config{uselongdouble} &&
+               ($Config{nvsize} == 16 || $Config{nvsize} == 12) &&
+               ($Config{longdblkind} == 3 ||
+                $Config{longdblkind} == 4));
     is(0x1p-1074,  4.94065645841246544e-324);
     is(0x1p-1075,  2.47032822920623272e-324, '[perl #128919]');
     is(0x1p-1076,  1.23516411460311636e-324);