browselist bug in Multibyte environment
Hiroshi MIURA
miura at samba.gr.jp
Wed Jul 5 16:46:43 GMT 2000
I've reciept a bug report in samba 2.0.6 and later.
The problem as follows:
He, the reporter, use a multibyte netbios name. In MS specification,
it is ok that we use multibyte netbios name on the NT and the 95/98.
In this case, coding system=euc and client code page =932.
But, now, he saw a MOJI BAKE (unintelligible sequence of characters)
on his 'Netowork Computer' on the Windows desktop.
I am starting read nmbd soruces, but I cannot understand the logic, yet.
I am trying to reappear it, and I get it. For investigating,
I look into the browse.dat, so it has two same name with diferrent
codes. When I set up coding sysmtem to EUC, so in the browse.dat
there are EUC(unix code) name and SJIS(windows native code) name.
That , I think, is because that some function in nmbd
forget a transform function call which convert sjis to euc
(unix_to_dos() or dos_to_unix()).
for example, I assume that
windows NT --------------> Samba(nmbd)
annouce by unicode uni->codepage code(sjis)
|
sjis to euc(depend coding system param.)
|
v
/var/samba/browse.dat(euc)
but in reverse data flow(for example),
windows NT <-------------- Samba(nmbd)
annouce by unicode uni <- codepage code(sjis)
but broken data ^ ^^^^^^ mis-match
| through (oh,bad)
|
/var/samba/browse.dat(euc)
then
windows NT --------------> Samba(nmbd)
annouce by unicode uni->codepage code(sjis)
about broken name | but broken sjis
sjis to euc(depend coding system param.)
| much worse.
v
/var/samba/browse.dat(euc)
I think that the problem is like this.
Do you have any idea?
--
SAMBA User Group Japan, devel. staff http://www.samba.gr.jp/
Linux Kernel Hack Japan, ProjectBLUE http://www.blue.gr.jp/
--- MIURA, Hiroshi http://www.momokuri.org/
More information about the samba-technical
mailing list