From: Junio C Hamano <gitster@pobox.com>
Date: Thu, 22 Jul 2010 17:35:35 GMT
Subject: Re: [PATCH] t2017: redo physical reflog existance check
Message-ID: <7vlj93h120.fsf@alter.siamese.dyndns.org>
In-Reply-To: <1279763190-32757-1-git-send-email-erick.mattos@gmail.com>

Erick Mattos <erick.mattos@gmail.com> writes:

> Although this is not a problem to all the other changes, it is laming to
> the tenth test.  The implementation of the correspondent problem creates
> a "touch" reflog that must be wiped out if not used by committing the
> new branch.

I thought about it a bit when I sent out my patch, but I do not think that
is necessary.

The things you care about, after running "-l --orphan eta", are:

 - If you make a commit, you get eta@{...} reflog that records it; and

 - If you leave the still-to-be-born eta branch without making a commit,
   you do not leave eta@{...} reflog behind.

Your zeta@{...} test is about the former, and your eta@{...} test is about
the latter.  I think they already check what they want to see happen.

I also am afraid that the "test -f" check would expose the implementation
detail more than necessary.  We may want to come up with a different
implementation of this behaviour later that may not create an empty file
there.