Re: git weird pulling issue

From: Junio C Hamano <gitster@pobox.com>
Date: Jan 3, 2008, 17:51 UTC
Message-ID: <7v4pdusspu.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <20080103121114.GE8046@cvg>

Cyrill Gorcunov <gorcunov@gmail.com> writes:

Show 24 quoted lines

> Hi git-list,
>
> i've a weird problem with pulling remote tree. look i've linus's tree
> as a base repo. then i've created x86 branch to which i pulled
> ingo's x86 tree. So all further pulling is made over this branch.
> And even having '--force' option at moment of pulling changes from
> Ingo's tree i've got something like that:
>
> ---
> cyrill@cvg linux-2.6-x86.git $ ./git-update.sh 
> Updating "x86"
> remote: Generating pack...
> remote: Done counting 15 objects.
> Result has 9 objects.
> remote: Deltifying 9 objects...
> remote: /9) done/9) done
> remote: Total 9 (delta 7), reused 3 (delta 1)
> Unpacking 9 objects...
>  100% (9/9) done
> Auto-merged include/asm-x86/msr.h
> CONFLICT (content): Merge conflict in include/asm-x86/msr.h
> Auto-merged include/linux/ptrace.h
> Auto-merged kernel/ptrace.c
> Automatic merge failed; fix conflicts and then commit the result.

To your readers, "./git-update.sh" is a mystery script; nobody knows what you are doing in there.

A handful points that might help anybody who wants to be helpful.

 * The state your repository is in is a bit vague.  Let me try
   to rephrase your introductory part and see if I understood
   your history.

    * It started as a clone from Linus's
      git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git

    * You say you created x86 branch (local) to which you pull
      Ingo's (I do not see any tree other than scheduler ones
      under git/mingo/), but it is unclear what this branch's
      history looks like.

      Was it done by

      	git checkout -b x86
	git pull git://.../mingo/x86/ master

      If so, (1) was the initial x86 before the first pull
      pristine Linus tree? (2) was the initial x86 pull
      fast-forward?

 * I am suspecting that git-update.sh is a wrapper around
   git-pull that does something like:

    git checkout x86 &&
    git pull git://.../mingo/x86/ master

   Do you have remote tracking information for Ingo's
   repository?  If so how does it look like?

   Do you have your own changes in your x86 branch?  Do they
   conflict with Ingo's change, IOW, is the conflict above just
   a normal thing to expect?  What makes you think there should
   not be any conflict?

There are a few reasons I can think of that can cause conflicts.

 - (obvious) you have your own changes that the tree you are
   pulling from have colliding changes;

 * the tree you are pulling from has rebased;

The --force is about allowing git-fetch to update the remote tracking branches (if you use them) even when the tracking information is not fast-forward. Without --force, the fetch would refuse to operate (and because git-pull is git-fetch followed by git-merge, the entire operation would fail and the merge won't even happen --- hence you would not even see conflicts). But the option does not change the fact that what you are going to merge has conflicting changes with the changes in your current branch. It merely allows git-fetch to forcibly update the remote tracking information and git-pull to proceed with the merges.

I am suspecting that (1) you have remote tracking information (perhaps set up with "git remote add"); and (2) Ingo's tree rebased between the time your last pull and this time. If you do not have any local commits on the branch, resetting to the remote tracking branch you use to track Ingo's progress is an option. If you do, then fetching and rebasing instead of pulling may make further development easier. But I cannot really tell.