HEAD works with Japanese? (Re: meeting with SUGJ)

[Andrew Tridgell]

Monyo wrote:
> At the meeting we examined only against Japanese Windows 2000, so I
> examined against Japanese Windows 98.
>
> With unix charset = UTF8, writing is almost successfull except around
> SFN that Andrew pointed out, but looking at the filenames from
> Windows, they are all broken though they are correctly written on
> UNIX filesystem.

I've just installed the japanese version of WinME in a vmware
session. I don't have japanese Win98 but I think WinME will be pretty
close.

You are right that WinME doesn't work very well with the Samba head
branch and Samba. Part of the problem is iconv, if you do this:

	echo '\' | iconv -f CP932 -t UTF8 

then you will see what the problem is. iconv in glibc 2.2.3 scews this
up badly, giving a multi-byte result for a 7 bit character! This
breaks tconx in Samba. 

Using libiconv-1.7 instead of the one in glibc is a better but I am
starting to get the impression that relying on iconv in Samba is a
mistake, especially given the unpredictability of the byte order of
the UCS2 charset in a given iconv implementation. We will really have
to support the ability to load either character tables or character
set plugins of some kind and I suspect these will be needed for good
Japanese support.

I'm currently on the plane on the way back to Australia. When I get
home I plan on creating a debian vmware system with Japanese fonts, so
I can see the problem you describe with the fonts on the unix side.

Cheers, Tridge