kaffe.io.ByteToCharIconv makes wrong results

[Ito Kazumitsu]

kaffe.io.ByteToCharIconv seems to make wrong results.

I use kaffe-snap-20010819.  In my environment, libiconv was not
automatically taken into kaffe,  so I did the following:
Afer ./configure, I edited config/config.h to add "#define HAVE_ICONV 1"
and edited libraries/clib/native/Makefile to add "-liconv" to LIBS.

Here is a simple program for testing.

import java.io.*;

public class TestIconv {
    public static void main(String[] args) {
        try {
            String from = args[0];
            InputStreamReader reader = new InputStreamReader(System.in, from);
            char[] cbuf = new char[1024];
            int l;
            while ((l = reader.read(cbuf,0,cbuf.length)) >= 0) {
                for(int i = 0; i < l; i++) {
                    System.out.print(cbuf[i]);
                }
            }
            System.out.flush();
        } catch (Exception e) {
            e.printStackTrace();
        }
    }
}

And here are the results of this program.

$ echo "aaaa" | kaffe TestIconv US-ASCII |od -xc
0000000 6161 0000 0000
          a   a  \0  \0  \0  \0
0000005
$ echo "aaaa" | kaffe TestIconv EUC-JP |od -xc
0000000 6161 0000 0000
          a   a  \0  \0  \0  \0
0000005
$ echo "aaaa" | kaffe TestIconv ASCII |od -xc
0000000 6161 6161 000a
          a   a   a   a  \n  \0
0000005
$ echo "aaaa" | kaffe TestIconv EUC_JP |od -xc
0000000 6161 6161 000a
          a   a   a   a  \n  \0
0000005
$

In the former two cases,  kaffe.io.ByteToCharIconv must have been
used to handle such encodings as "US-ASCII" and "EUC-JP".  And
the results are wrong.

In the latter two cases, kaffe.io.ByteToCharASCII and
kaffe.io.ByteToCharEUC_JP must have been used.  And the results
are satisfactory.

This comparison makes me think there is some problem around
kaffe.io.ByteToCharIconv.