gitweb/test/Marchen always untracked?

I'm not sure what is going on here but on my Mac OS X system
the file `gitweb/test/Marchen` is always untracked:

  [spearce@spearce-pb15 git]$ git status
  # On branch refs/heads/sp/work
  #
  # Untracked files:
  #   (use "git add" to add to commit)
  #
  #       gitweb/test/Märchen
  nothing to commit
  
  [spearce@spearce-pb15 git]$ git-ls-files | grep chen
  "gitweb/test/M\303\244rchen"
  
  [spearce@spearce-pb15 git]$ ls -l gitweb/test/Ma\314\210rchen 
  -rw-r--r--   1 spearce  spearce  17 Jul  8 14:38 gitweb/test/Ma??rchen

At first glance it would appear as though the file is encoded into
the tree (and thus the index) one way and the OS is reporting it
another way during a readdir.  Which is leaving me scratching my
head...

-- 
Shawn.

Hi,

On Sat, 8 Jul 2006, Shawn Pearce wrote:

I'm not sure what is going on here but on my Mac OS X system
the file `gitweb/test/Marchen` is always untracked:

I noticed that problem also. It seems to be yet another encoding problem, 
because you see the Latin1 interpretation of a UTF-8 encoded umlaut, ä.

What is fascinating:

git-ls-files --cached|grep chen.$

"gitweb/test/M\303\244rchen"

but

git-ls-files --others|grep chen.$

"gitweb/test/Ma\314\210rchen"

Will investigate,
Dscho

Hi,

On Sat, 8 Jul 2006, Johannes Schindelin wrote:

On Sat, 8 Jul 2006, Shawn Pearce wrote:

quoted

I'm not sure what is going on here but on my Mac OS X system
the file `gitweb/test/Marchen` is always untracked:

Will investigate,

And so I did. It seems the problem is not reading, but _writing_.

touch $(echo -e 'M\xc3\xa4rchen')
ls *en|hexdump -C

00000000  4d 61 cc 88 72 63 68 65  6e 0a                    |Ma..rchen.|
0000000a

So, instead of writing "4d c3 a4 72 63 68 65 6e" it actually writes "4d 61 
cc 88 72 63 68 65 6e". It might be possible that this is some 
"intelligent" encoding mapping, but ...

echo $LC_ALL

de-DE.utf-8

... so I think Mac OS X is wrong.

Ciao,
Dscho