Don't use utf8.pm in the test (avoiding embedding utf8 in the test

file), and use greek to test for something out of 8-bit latin, as
Jarkko suggested. Everything flows.

p4raw-id: //depot/perl@30805

diff --git a/t/io/dup.t b/t/io/dup.t
index 0287023..d101688 100755 (executable)
--- a/t/io/dup.t
+++ b/t/io/dup.t
@@ -133,10 +133,10 @@ SKIP: {
     }
     close G;
 
-    use utf8;
     open UTFOUT, '>:utf8', "dup$$" or die $!;
     open UTFDUP, '>&UTFOUT' or die $!;
-    my $message = "ça marche pas\n";
+    # some old greek saying.
+    my $message = "\x{03A0}\x{0391}\x{039D}\x{03A4}\x{0391} \x{03A1}\x{0395}\x{0399}\n";
     print UTFOUT $message;
     print UTFDUP $message;
     binmode UTFDUP, ':utf8';
@@ -151,7 +151,6 @@ SKIP: {
        $line = <UTFIN>; is($line, $message);
     }
     close UTFIN;
-    no utf8;
 
     END { 1 while unlink "dup$$" }
 }