Here is an example:

preparation: put an umlaut into a file and gzip it. Also

echo ü > umlaut
gzip -k umlaut
[download]

Now check the difference :(

perl -e '
  use IO::Uncompress::Gunzip;
  binmode(STDOUT, ":utf8");
  open my $in, "<:utf8", "umlaut";
  $_=<$in>;
  print "Uncompressed: $_ ",ord($_),"\n";
  my $gin= IO::Uncompress::Gunzip->new("umlaut.gz");
  binmode($gin, ":utf8");
  $_=<$gin>;
  print "  Compressed: $_ ",ord($_),"\n";
'
[download]

Output

Uncompressed: ü
 252
  Compressed: ü
 195

In theory there shouldn't be a difference between the outputs :(

Update: I learned that "binmode" won't do anything to the IO::Uncompress::Gunzip filehandle.

Handling the decode myself, not relying on an IO-layer, gives the expected result:

perl -e '
  use IO::Uncompress::Gunzip;
  use Encode;
  binmode(STDOUT, ":utf8");
  my $gin= IO::Uncompress::Gunzip->new("umlaut.gz");
  $_=<$gin>;
  $_ = Encode::decode("UTF-8", $_);
  print "  Compressed: $_ ",ord($_),"\n";
'
[download]

Update: As suggested by Corion I'm now using PerlIO::gzip. My Original code, note the test example shown here, now is:

        my $encoding = ":utf8";
        if ( $filename =~ /\.gz$/ ) {
            $encoding = ":gzip$encoding";
        }
        open $in, "<$encoding", $filename or die "Can't read $filename
+: $!\n";
[download]