Hi,
On Mon, 10 Mar 2008, Sebastian Schuberth wrote:
> I'm having several autocrlf related issues with the new Git-
> preview20080301.exe installer.
>
> The first is rather an oddity than a real problem:
>
> $ grep autocrlf ~/.gitconfig
> autocrlf = true
>
> $ grep autocrlf .git/config
>
> $ git config --get core.autocrlf
> true
>
> $ git config --list | grep autocrlf
> core.autocrlf=true
> core.autocrlf=true
>
> As you can see, "config --list" lists autocrlf twice. Probably the new
> internal default of autocrlf=true is listed here too?
It comes from /etc/gitconfig, yes.
> The second problem in fact seems to have already existed in earlier
> releases, but now became more apparent due to the warning Git is
> issuing:
>
> $ git config --get core.autocrlf
> true
>
> $ git branch
> vi-stable
> * vi-work
>
> $ git merge origin/qt4
> warning: LF will be replaced by CRLF in amiramain/bin/start
> Auto-merged Amira/HxObject.cpp
> Auto-merged hxcolormapeditor/HxAdvancedColormapEditor.cpp
> CONFLICT (content): Merge conflict in hxcolormapeditor/
> HxAdvancedColormapEditor.cpp
> Automatic merge failed; fix conflicts and then commit the result.
>
> # Both the "start" and "HxAdvancedColormapEditor.cpp" have Unix line-
> ending at this point,
> # despite autocrlf=true
>
> $ git mergetool
> Merging the files: hxcolormapeditor/HxAdvancedColormapEditor.cpp
>
> Normal merge conflict for 'hxcolormapeditor/
> HxAdvancedColormapEditor.cpp':
> {local}: modified
> {remote}: modified
> Hit return to start merge resolution tool (ecmerge):
> warning: LF will be replaced by CRLF in hxcolormapeditor/
> HxAdvancedColormapEditor.cpp
>
> # Again, the warning is correct, but HxAdvancedColormapEditor.cpp
> should have had CRLF instead
> # of LF in the first place
>
> Moreover, if there is a modified file
>
> $ git diff --name-status
> M hxvoltex/SxVoltex.h
>
> then "git diff" will checkout the temporary file which reflects
> "SxVoltex.h" in the repository also with LF instead of CRLF, i.e. the
> autocrlf setting does not seem to be respected. But this problem also
> was present in earlier releases already.
>
> Steffen Prohaska already took a quick look at this (but has no time to
> fix it) and thinks handle_file() in builtin-rerere.c is missing the
> conversion and thus causing the problems.
I hope to get some time and motivation tonight to have a look at it, could
you bug me again tomorrow if I did not?
Ciao,
Dscho
