Re: HFS+ Unicode weirdness

From: David Symonds <dsymonds@gmail.com>
Date: Sep 4, 2007, 13:07 UTC
Message-ID: <ee77f5c20709040607u163d1336k46c83cfb9e37bfee@mail.gmail.com>
In-Reply-To: <052099D2-F79B-4063-82D3-BFB5D0102A55@wincent.com>

On 04/09/07, Wincent Colaiuta <win@wincent.com> wrote:

Show 11 quoted lines

> On a brand new clone of git.git the file "gitweb/test/Märchen" is
> provoking some weird behaviour running on Mac OS X and the toy HFS+
> filesystem. Note how the unmodified checkout of the file is shown as
> "untracked" by "git st", but on deleting the file it's shown as
> "deleted". If I build a copy of Git based on the clean working tree
> then the resulting build has a version number of "1.5.x-dirty".
>
> Any suggestions on where to start investigating the cause of this?
> About the only lead I have is that if I create a file with that name
> by typing it's name  it's encoded as "Ma\314\210rchen", but the file
> in the git.git repo is encoded as "M\303\244rchen".

It's because OS X is using composite Unicode encoding (storing "ä" as "a" + umlaut). Git doesn't canonicalise things, since OS X does its jiggery-pokery of that kind of stuff in CoreFramework, which I believe Git doesn't use.

Dave.