Problems for Japanese support on Samba 3.0

[TAKAHASHI Motonobu]

Jelmer Vernooij wrote:
>>But anyway using Samba 3.0 under Japanese environment, to support
>>at least SJIS and EUC correctly is required and it's better to support
>>them in libiconv. I will ask Bruno Haible <bruno at clisp.org>, the
>>author of libiconv, again to modify/add Japanese character sets.
>>If he didn't reply, I will think of implementing as a module.
>Ok.

There are nothing to reply...

Anyway, I checked Samba 3.0alpha23 and found 2 problems.

- toupper'ed part of 2byte chars: for example
  under "unix charset = CP932", a longname for a character which code
  is 0x8161 returned via GetLongPathName() is changed to 0x8141.

- a character which contained a byte of 0x5c will be mangled: for
  example, under "unix charset = CP932", a filenames for a character
  which code is 0x817c (or 0x835c) is mangled, but these should not be
  mangled.

I put the result at 
  http://www.monyo.com/technical/samba/jchartest/20030505/

please diff (or windiff) between LOCAL.TXT 
  and samba-3.0alpha23-CP932-20030505.TXT.
LOCAL.TXT is the result against to Windows XP machine.

And Japanese Windows NT(also in 2000/XP) has unique case insensitive
feature for some KANJI characters which mean roman numeric (for
example, a 2byte character for 'IV'(U+2163) and a 2byte character
'iv'(U+2173) should be a same character as 'A' and 'a'. Where to put
this routine in current Samba code?

-----
TAKAHASHI, Motonobu (monyo)                    monyo at home.monyo.com
                                               http://www.monyo.com/