Re: git-cvsimport bug with dates

From: Elijah Newren <newren@gmail.com>
Date: Nov 17, 2007, 15:39 UTC
Message-ID: <51419b2c0711170739p1a2bf6c7g87f615676f72aedf@mail.gmail.com>
In-Reply-To: <200711171112.23150.robin.rosenberg.lists@dewire.com>

On Nov 17, 2007 3:12 AM, Robin Rosenberg <robin.rosenberg.lists@dewire.com> wrote:

Show 17 quoted lines

> fredag 16 november 2007 skrev Elijah Newren:
>
> > On Nov 15, 2007 11:06 PM, Junio C Hamano <gitster@pobox.com> wrote:
> > > When you use "branch@{date}" notation, you are not asking a
> > > question on the project history, but a question on the local
> > > view from _your_ repository on that project.
> >
> > Interesting; that makes sense from a merge or pull viewpoint, but
> > wouldn't it make more sense to have cvsimport ensure the commits are
> > treated as though they actually existed in master as of the date
> > specified in CVS?
>
> Reflog do not work that way. They don't say when a commit entered a repo,
> only when a ref changed. For a CVS import things could work as you suggest
> but I think the confusion among newcomers would be massive if people start
> using reflogs  'as if' it said anyting about when a commit entered. It can be used
> as a hint.

Okay...so I guess my suggestion could be translated as: don't worry about when the commit entered; instead make git-cvsimport mark the refs as changed as of the CVS date mentioned instead of as of the import time.

I guess there would be reluctance to do this whenever git-cvsimport would be used for incremental commits on a repository that also had local commits? Sounds like the issue is deeper than I first realized and this probably isn't worth the effort it'd take.

Thanks for the explanations.

Elijah