Some Japanese characters cannot be displayed in the Japanese locale with JRE 1.3.1_02 using JIS0208 character encoding.  The glyphs for the characters that fail to display are available on the Japanese Solaris as the characters are easily displayed on a dtterm window.  It seems that the JRE is missing a mapping of these characters to the correct font/glyph.

A patched version of CharToByteJIS0208.java has been created which fixes the problem.  The problem also does not exist on 1.4.x.  

To reproduce the problem, you need the Japanese language support (EUC encoding). Next, you need to make sure that the terminal you use is set up such that the locale command will display the following:

% locale
LANG=ja
LC_CTYPE="ja"
LC_NUMERIC="ja"
LC_TIME="ja"
LC_COLLATE="ja"
LC_MONETARY="ja"
LC_MESSAGES="ja"
LC_ALL=

Next, cat the attached file "characters".  On your terminal, you should see something like " customer  Word - VIII".  Please select the Roman numerals V and  customer  by using the Sun copy key.   

Next, do the following:  

%cd /usr/demo/J2SE/demo/jfc/Stylepad
% java -jar Stylepad.jar &

Once the application comes up, use the Sun paste key to paste what you just copied from the characters file.  The characters will be displayed as square boxes.  

If you replace CharToBytesJIS0208.java with the attached modified version, the characters are displayed properly.  The diffs between the original and modified files are also attached.

The attached fonts.dir file shows the fonts used in the application.  The font has support for additional Unicode glyphs than those in CharToByteJIS0208.java.  The modifications to the file simply add the additional mappings that the font (HG-MonchoL.ttf- customer -hg mincho l-medium-r-normal--0-0-0-0-m-0-jisx0208.1983-0) is capable of handling.

N/A

Currently, Java uses the eucJP converters which are compliant with the Japanese locale standard provided by Japanese UNIX vendors (the result of Open Group activities). However, Solaris supports its proprietary extensions to the standard. See the eucJP(5) man page.

The fix should be to provide Solaris eucJP compatible converters and make them default converters on Solaris for the ja locale. We shouldn't change the JIS X 0208 converter tables. The JISAutoDetect converter needs to use the Solaris proprietary converter (similar to SJIS/MS932 setup) for "eucJP". New converters for the Solaris PCK are also required. PCK is not SJIS. See the PCK(5) man page. This requires additional change to JISAutoDetect for SJIS/MS932/PCK setup. 
  xxxxx@xxxxx   2002-10-21


  xxxxx@xxxxx   2002-10-23

Mapping of so called "NEC (row 13) special charcters (or symbols)"
and "IBM extended characrers" are included in the conversion policy
provided by The Open Group Japan Vendor Council (TOG JVC),
which looks to be identical to what   xxxxx@xxxxx   is refering to.
i.e. the mapping is not Sun or Solaris proprietary extension.

See: http://www.opengroup.or.jp/jvc/cde/appendix-e.html

eucJP(5) refers to the characters as VDCs (vendor defined
characters) because they are not defined by JIS itself.

========================================================================
Fix in progress (and under review) for 1.3.1_x
Also NIO coders for the modified Solaris mappings are also being
prepared - ie. JIS_X_0208_Solaris, JIS_X_0212_Solaris, EUC_JP_Solaris
and PCK

  xxxxx@xxxxx   2002-11-05

Based on further information provided from i18n/Japan Solaris group some of the
originally supplied roundtrip mappings were not correctly implemented.
I have made the following changes:

Before:   0x2d75(JIS X 0208) <--> U+221A, 0x2265 (JIS X 0208) ---> U+221A
Now:      0x2d75(JIS X 0208) ---> U+221A, 0x2265 (JIS X 0208) <--> U+221A
Before:   0x2d77(JIS X 0208) <--> U+2220, 0x225C (JIS X 0208) ---> U+2220
Now:      0x2d77(JIS X 0208) ---> U+2220, 0x225c (JIS X 0208) <--> U+2220
Before:   0x2d7c(JIS X 0208) <--> U+222A, 0x2240 (JIS X 0208) ---> U+222A
Now:      0x2d7c(JIS X 0208) ---> U+222A, 0x2240 (JIS X 0208) <--> U+222A
Before:   0x2d72(JIS X 0208) <--> U+222B, 0x2269 (JIS X 0208) ---> U+222B
Now:      0x2d72(JIS X 0208) ---> U+222B, 0x2269 (JIS X 0208) <--> U+222B
Before:   0x2d71(JIS X 0208) <--> U+2261, 0x2261 (JIS X 0208) ---> U+2261
Now:      0x2d71(JIS X 0208) ---> U+2d71, 0x2261 (JIS X 0208) <--> U+2261
Before    0x2237(JIS X 0212) ---> U+007E
Now:      0x2237(JIS X 0212) <--> U+FF5E

The supplied converters are now implemented similar to the Solaris iconv 
converters. This results in updates in the char->byte mappings for the
Solaris EUC_JP and PCK converters being added as part of addressing this bug.

  xxxxx@xxxxx   2002-12-11
  xxxxx@xxxxx   2002-12-11

removed 1.4.1_04 from commit to fix field as part of bug clean-up.
  xxxxx@xxxxx   2004-09-16