Re: [PATCH v2 2/2] git-merge-file: do not add LF at EOF while applying unrelated change

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Jul 2, 2014, 14:08 UTC
Message-ID: <alpine.DEB.1.00.1407021604510.14982@s15462909.onlinehome-server.info>
In-Reply-To: <20140702044427.GA7325@wheezy.local>

Hi Max,

On Wed, 2 Jul 2014, Max Kirillov wrote:

> On Mon, Jun 30, 2014 at 04:55:10PM +0200, Johannes Schindelin wrote:
> > I just wish the tests were a little easier to understand...
> 
> What could be improved with them?

Oh, I would name the files more appropriately, for example. That is, instead of test1.txt I would call it mixed-endings.txt or lf-only.txt or some such.

And instead of the Latin version of Psalm 23, I would put lines into the files that describe their own role in the test, i.e.

	unchanged
	ends with a carriage return
	ends with a line feed
	unchanged

or similar.

Please keep in mind that this critique is most likely on my *own* work, for all I know *I* introduced those files.

Show 6 quoted lines

> By the way, for "\r\n" eol it did even worse, adding just "\n". And I
> guess it still adds just "\n" for union merge.  Should file merge
> consider the core.eol? I think it should, and for the conflict markers
> also, it looks ugly when whole file has "\r\n" but the conflict markers
> have "\n". But then git-merge-file could not be used outside of
> repository, I guess.

Oh, why not? It could read the configuration if it's inside a working directory, and just read /etc/gitconfig and $HOME/.gitconfig when outside...

> In general, I wish file merging (and diffing) were more tolerant of the
> line endings in input. Because in windows environment, when people have
> different core.autocrlf, it becomes quite frustrating to always get
> conflicts and changes.

Amen!

Ciao, Dscho