2007/4/20, Tony Wu <[EMAIL PROTECTED]>:
Hi all,
There is a difference in default system property - "file.encoding". It
comes from vm[1]. I considerd it was a non-bug difference because
there isn't any description in spec. What's your opinion?
[1]
e.g. for locale zh_CN, the output
IBM VM: GB18030
DRLVM: 8859-1
RI : GBK
3307 and 3702 are related issues.
There are several open entries in SUN's bug parade for documenting
"extra" properties, typical resolution says [1]:
"Users are discouraged from any reliance on implementation details.
The key "file.encoding" in particular is an implementation artifact
that may not be defined or used by other vendors."
But since we rely on this property in *our implementation*, we need to
perform flawlessly - putting aside the difference with RI.
DRLVM simply hardcodes this property value, no any runtime detection
is performed. I think it would be nice if classlib could set up
appropriate value at startup (new JIRA needed?).
[1] http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6334530
---------- Forwarded message ----------
From: Vasily Zakharov (JIRA) <[EMAIL PROTECTED]>
Date: Apr 20, 2007 2:46 PM
Subject: [jira] Commented: (HARMONY-3702) [classlib][luni] Reader and
Writer convert characters incorrectly
To: [EMAIL PROTECTED]
[
https://issues.apache.org/jira/browse/HARMONY-3702?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12490241
]
Vasily Zakharov commented on HARMONY-3702:
------------------------------------------
I'm not sure if this is a non-bug difference. Probably we should have
the same default encoding as RI, for compatibility.
Probably this should be discussed in the mailing list.
> [classlib][luni] Reader and Writer convert characters incorrectly
> -----------------------------------------------------------------
>
> Key: HARMONY-3702
> URL: https://issues.apache.org/jira/browse/HARMONY-3702
> Project: Harmony
> Issue Type: Bug
> Components: Classlib
> Reporter: Vasily Zakharov
> Attachments: test.dat
>
>
> java.io.Reader converts bytes to characters differently than RI does. Also,
java.io.Writer converts characters to bytes differently than RI does.
> The attached test.dat file contains random test data and must be placed to
the current directory. ReaderTest below reads that file with FileReader and then
dumps it to standard output by converting each character to int. WriterTest reads
the test.dat file with FileInputStream, converts each byte to character by casting
and then dumps the resulting characters to standard output by OutputStreamWriter.
> public class ReaderTest {
> public static void main(String args[]) throws Exception {
> char[] buffer = new char[0x100000];
> java.io.Reader reader = new java.io.FileReader("test.dat");
> int length = reader.read(buffer, 0, buffer.length);
> for (int i = 0; i < length; i++) {
> System.out.println((int) buffer[i]);
> }
> }
> }
> public class WriterTest {
> public static void main(String args[]) throws Exception {
> byte[] buffer = new byte[0x100000];
> java.io.InputStream iStream = new java.io.FileInputStream("test.dat");
> int length = iStream.read(buffer, 0, buffer.length);
> char[] charBuffer = new char[length];
> for (int i = 0; i < length; i++) {
> charBuffer[i] = (char) buffer[i];
> }
> java.io.Writer writer = new java.io.OutputStreamWriter(System.out);
> writer.write(charBuffer, 0, length);
> writer.close();
> }
> }
> In both cases, output files on RI and on Harmony are different:
> $ RI/bin/java ReaderTest > reader.ri
> $ HY/bin/java ReaderTest > reader.hy
> $ diff --binary -q reader.ri reader.hy
> Files reader.ri and reader.hy differ
> $ RI/bin/java WriterTest > writer.ri
> $ HY/bin/java WriterTest > writer.hy
> $ diff --binary -q writer.ri writer.hy
> Files writer.ri and writer.hy differ
> My investigations show that the problem is in Reader/Writer, not in
InputStream/OutputStream. Also, I've tried other implementations of Reader/Writer
and they share the same problem.
> The problem was discovered on Windows XP/IA-32 but probably affects other
platforms too.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
--
Tony Wu
China Software Development Lab, IBM
