rsync - Mar 2006 - rsync plus cygwin: language problems with filenames

I know this problem has been posted before, but I have been unable to 
find a solution on this list or anywhere else on the net.  Let me review.

First, here is my setup:
OS: Windows XP Pro SP2
Rsync: 2.6.6-1
Cygwin: 2.510.2.2

Whenever I transfer any file that has Japanese characters in it, the 
following message appears in the screen output:

C:\Documents and Settings\Steve>rsync -avvz --delete /cygdrive/c/test 
/cygdrive/c/test2
building file list ...
file has vanished: "/cygdrive/c/test/01 - ??????????.mp3"
done
deleting in test
IO error encountered -- skipping file deletion
delta-transmission disabled for local transfer or --whole-file
test/New Text Document.txt is uptodate
total: matches=0  tag_hits=0  false_alarms=0 data=0

sent 119 bytes  received 32 bytes  302.00 bytes/sec
total size is 0  speedup is 0.00
rsync warning: some files vanished before they could be transferred 
(code 24) at
 /home/lapo/packaging/tmp/rsync-2.6.6/main.c(791)

So far, this is the best explanation I have found for what may be happening:
http://lists.samba.org/archive/rsync/2005-May/012599.html

Also, people have pinned it down to a Windows language setting problem:
http://lists.samba.org/archive/rsync/2005-May/012611.html

I confirmed this for myself by going into Regional and Language Options, 
Advanced, and choosing Japanese for the language I wanted to use in 
non-Unicode programs.  Now, it will transfer the Japanese files fine.  
It will also transfer most English language files.  However, it then has 
a problem with accented letters, which I assume must have been part of 
the English (United States) settings.

I further proved that it is very likely just a Windows problem by trying 
to copy one of the files in Ubuntu Linux.  It worked perfectly with 
rsync 2.6.5 and I bet it would've done the same thing with 2.6.6.

Has anyone made any progress in solving this problem?  I noticed that 
the NEWS file for rsync 2.6.7 
(http://www.samba.org/ftp/rsync/rsync-2.6.7-NEWS) discusses an 
improvement in escaping unreadable characters.  Is that related to this 
problem?  I'm new to cygwin and I'm not sure which binary I should be 
using or whether I have to wait for cygwin to come out with it in their 
own system.

I'd really like to get this to work.  This is one tiny problem in an 
otherwise flawless solution to my backup needs.  Please let me know if 
you have any ideas to make this work.

Thanks for reading.

-Steve

I thought the problem was that cygwin doesn't support unicode.

See here:
http://lists.samba.org/archive/rsync/2002-September/003643.html

The post is rather old, but for whatever reason the cygwin folks
haven't added unicode support.  You'll also see this problem manifest
itself as a limit in the path length that the cygwin compiled version
of rsync supports.

On Mon, Mar 13, 2006 at 11:12:48PM -0800, Steve Robinson
wrote:
> I further proved that it is very likely just a Windows problem by trying 
> to copy one of the files in Ubuntu Linux.
Yes, this is something that MS Windows is doing outside of rsync's
control, so it is not something that the main rsync distribution is even
looking at because there's nothing to fix in the main distribution.
This may be something that the cygwin folks can look into fixing,
though.

The problem is that rsync asks for the name of a file via readdir(), and
MS Windows munges the name instead of returning it verbatim.  Rsync then
tries to use the provided filename to open the file, and MS Windows says
that it doesn't exist.  At that point there's nothing rsync can do.

Perhaps the cygwin version of rsync could be marked as understanding
unicode -- something that I assume the cygwin packagers would do when
building the rsync package.

..wayne..