Re: [PATCH/RFC] git-fetch: Only fetch in fetch

From: Junio C Hamano <junkio@cox.net>
Date: Jan 25, 2007, 01:13 UTC
Message-ID: <7vhcufc3j5.fsf@assigned-by-dhcp.cox.net>
In-Reply-To: <8aa486160701241652x375f0d0eoba1cdf3d14550034@mail.gmail.com>

"Santi Béjar" <sbejar@gmail.com> writes:

Show 13 quoted lines

> On 1/25/07, Santi Béjar <sbejar@gmail.com> wrote:
>>    I've tried to separate the fetch and merge logic. At the same time it
>>    has removed 40 lines of code to the git-*.sh.
>
> Just some words for the basic idea.
>
> git-fetch: fetches the branches from the remote and save this
> information in .git/FETCH_HEAD. In .git/FETCH_REMOTE it save
> additional information, as the remote nick, the remote url, the
> refspec used and the default remote.
>
> git-pull: uses this information to select the branches to merge and
> save them in .git/FETCH_MERGE.

One thing I suspect is problematic is that this would make fmt-merge-message much less useful after running git-fetch. You have to go all the way to git-pull and replace reading from FETCH_HEAD with reading from FETCH_MERGE (just like you had to do that in git-pull). In other words, this may break people's scripts and workflows.

This is probably a good clean-up at the conceptual level. If you had done this before v1.0.0 there would have been no question about its inclusion.

Please discuss this patch post 1.5.0.