** Tested with LocaleSelftest for Win32 (4/9/99-#2) at /u/momoi/Sort/win. on 4/22/99 Win32 5.0 build ** ** Conducted under NT4-Japanese ** I used a primitive Japanese sort data found above called "jpnsorttemp.txt" and applied LocaleSelfTest on it with and without the "ja_JP" locale option. The results were the same. Kana and Kanji were mixed in sorted order and no discernible princiople could be detected. It seems then that the test is not reliable for Japanese yet.

This was because of a lack of unicode conversion in the test program. I fixed it for windows, latest binary was copied to /u/momoi/Sort/win There is a known problem of collation and date/time format. In Win95 or Win98, it only works for Latin1 (need an additional service in nsLocale). This won't be fixed before M5 so do the testing using NT only.

** Checked with 4/26/99 Win32 build ** With the 4/23/99 version of LocaleSelfTest.exe, the above primitive test file now show correct ordering. Given this result, we should now be able to test the full Shift_JIS sort data with the test program. If I find problems with the full date, I'll open another bug. This bug was about a very basic sor order issue, which nhotta's fix seemed to have resolved. I'm marking this fix verified.