# Cygwin: Initial clone of repo containing .gitattributes has modified files

3 messages from 2009-06-02 to 2009-06-02. Participants: Raman Gupta, Junio C Hamano.
Thread: https://gitlist.dev/t/19636

## Raman Gupta, 2009-06-02 12:47

Subject: Cygwin: Initial clone of repo containing .gitattributes has modified files
Message-ID: <4A251F66.2020809@fastmail.fm>
URL: https://gitlist.dev/e/4A251F66.2020809%40fastmail.fm

```
When a repository containing a .gitattributes file, which contains
other files with -crlf set is cloned on cygwin, with autocrlf = true,
the initial checkout of the working copy sets the line endings for the
-crlf files incorrectly (it ignores the .gitattributes, presumably
because it hasn't been checked out yet).

After the checkout has been done, git does correctly mark the files
with -crlf as "modified" and a simple reset --hard fixes it, but this
is very confusing for the user as the user has not done anything -- it
is git itself that has "modified" the file by not assigning the
correct line endings.

Reproduction recipe (on a cygwin binmode mount):

---------------------------------------------------
$ git --version
git version 1.6.1.2
$ git config --global core.autocrlf true
$ git config --global core.safecrlf true
$ mkdir temp.git; cd temp.git; git init
$ cat | unix2dos > file_dos
DOS
line
endings
^D
$ cat > file_unix
UNIX
line
endings
^D
$ cat > .gitattributes
.gitattributes -crlf
file_unix -crlf
^D
$ git add .gitattributes *
$ git commit -m "Test"
$ cd ..
$ git clone temp.git tempclone.git
$ cd tempclone.git
$ git status
...
#      modified:  .gitattributes
#      modified:  file_unix
...
---------------------------------------------------

Also, in a related issue, git on cygwin requires .gitattributes itself
to be included in .gitattributes, otherwise with safecrlf = true you get:

$ git add .gitattributes
fatal: LF would be replaced by CRLF in .gitattributes

(since .gitattributes created with a cygwin editor on a binmode mount
uses unix line endings)

Cheers,
Raman Gupta

```

## Junio C Hamano, 2009-06-02 15:50

Subject: Re: Cygwin: Initial clone of repo containing .gitattributes has modified files
Message-ID: <7vd49mfxcv.fsf@alter.siamese.dyndns.org>
URL: https://gitlist.dev/e/7vd49mfxcv.fsf%40alter.siamese.dyndns.org
In-Reply-To: <4A251F66.2020809@fastmail.fm>

```
Raman Gupta <rocketraman@fastmail.fm> writes:

> When a repository containing a .gitattributes file, which contains
> other files with -crlf set is cloned on cygwin, with autocrlf = true,
> the initial checkout of the working copy sets the line endings for the
> -crlf files incorrectly (it ignores the .gitattributes, presumably
> because it hasn't been checked out yet).

I think this is what v1.6.3-rc0~83^2~1 (Read attributes from the index
that is being checked out, 2009-03-13) was about.

I see you are using a version based on 1.6.1.2; isn't there a newer Cygwin
packaged one available?

```

## Raman Gupta, 2009-06-02 15:56

Subject: Re: Cygwin: Initial clone of repo containing .gitattributes has modified files
Message-ID: <4A254BA6.3080601@fastmail.fm>
URL: https://gitlist.dev/e/4A254BA6.3080601%40fastmail.fm
In-Reply-To: <7vd49mfxcv.fsf@alter.siamese.dyndns.org>

```
Junio C Hamano wrote:
> Raman Gupta <rocketraman@fastmail.fm> writes:
> 
>> When a repository containing a .gitattributes file, which contains
>> other files with -crlf set is cloned on cygwin, with autocrlf = true,
>> the initial checkout of the working copy sets the line endings for the
>> -crlf files incorrectly (it ignores the .gitattributes, presumably
>> because it hasn't been checked out yet).
> 
> I think this is what v1.6.3-rc0~83^2~1 (Read attributes from the index
> that is being checked out, 2009-03-13) was about.

Thanks! I looked for the fix but didn't find any reference to it. Sorry.

> I see you are using 

Actually, I use Linux -- its my coworkers that use cygwin! Bleh! :)

> a version based on 1.6.1.2; isn't there a newer Cygwin
> packaged one available?

Nope. That is the latest available as of this morning (at least in the
cygwin non-experimental branch).

Cheers,
Raman

```