I need an advice

Wed Aug 15 23:59:00 GMT 2001

Hi!
I investigated my problem from
http://gcc.gnu.org/ml/java/2001-05/msg00110.html
a bit deeper and found out this:
(Sorry for the delay, but debugging on that machine is no fun since
a 'gmake && gmake install' lasts for more than half an hour).
My application reads a binary file, transforms the data and builds
String objects by 'new String' out of a byte Array.
This calls 
public String (byte[] byteArray, int offset, int count)
from java/lang/String.java
and tries init with the default file encoding from the system property
"file.encoding".
I did not set anything special here, so my Solaris box responds to 
$ locale
with
LANG=
LC_CTYPE="C"
LC_NUMERIC="C"
LC_TIME="C"
LC_COLLATE="C"
LC_MONETARY="C"
LC_MESSAGES="C"
LC_ALL=
The system property "file.encoding" is thus set to "646" in natSystem.cc
This encoding seems to be either incorrect or it cannot be handled,
so i get an exception each time i use new String(byteArray) from my
application. This occures 180000 times and each time i use it, 
java::lang::String::init (jbyteArray bytes, jint offset, jint count,
 jstring encoding)
from natString.cc tries
gnu::gcj::convert::BytesToUnicode::getDecoder(encoding);
with this wrong or not correctly handled encoding and throws an exception,
because getDecoder's command
decodingClass = Class.forName(className)
does not result in anything usable.
It just creates traffic to my harddisk :-(
Of course the same problem applies the other way round (String --> ByteArray)
So i have some options:
- The quick fix for me was 
 #define DEFAULT_FILE_ENCODING "8859_1" instead of
 #define DEFAULT_FILE_ENCODING file_encoding ()
 in natSystem.cc
 That's not nice, but it works.
- Set LC_??? to anything useful; i tried a lot of possibilities but nothing
 worked (neither UTF-8 nor 8859_1 in LC_ALL or LC_CTYPE)
- Check inside natSystem.cc, if the file.encoding actually works and replace 
 it by "8859_1" inside file_encoding() if not.
 So we always have a working fall back. Does this conform to the java docs?
- Cache the last working encoding in 
 public String (byte[] byteArray, int offset, int count)
 inside String.java.
 Nevertheless we must make sure, that changes during
 runtime of "file.encoding" get recognized.
- Do the same in natString.cc
 java::lang::String::init (jbyteArray bytes, jint offset, jint count,
 jstring encoding)
 with encoding.
Any words of wisdom here?
Thanks
Martin.
PS: This is all tested using the gcc-20010813 snapshot on a Solaris 5.7 box.
-- 
The early bird catches the worm. If you want something else for 
breakfast, get up later.