Re: Question on multi-level git repository heiarchy.

[Andy Parkins <andyparkins@gmail.com>]

On Friday 2006 October 27 15:05, Ben Greear wrote:

> I want to synchronize the entire git repo, including all branches.  How
> would I go
> about doing that?

That happens anyway.  "git pull" is actually a fetch + a merge.  Personally I 
recommend never doing a "git pull" without arguments.  Instead:

First, update every "Pull:" line from .git/remotes/origin
 $ git fetch
Now manually merge one of the (now locally stored) upstream trees into your 
working tree.
 $ git pull . upstream-tree

> Is there any way to recover my currently mangled tree on the
> workstation, or do I need
> to start fresh there?  If I start fresh, do I re-clone, or is there some
> better way to get

I'm relatively new to git; my experience is that there is very little that 
can't be undone.  Exactly what sort of mangling has happened to your tree?

If you've got yourself in a real state, I suggest editing .git/remotes/origin 
to pull the upstream trees to some newly named branches (maybe just 
add "new/" to the front of them).  Treat those as read-only (i.e. don't check 
them out), and start comparing branches.  "gitk --all" is incredibly useful 
for this.

Make yourself a new branch for your own work based off a new/upstream branch  
(git checkout -b fixthings new/upstream), then use git-cherry-pick to pull 
your patches out of your mangled tree to build up a branch with your changes 
on it.

Now, you can delete all the branches that you think are mangled and (if you 
really want to), rename all the new/ refs back to whatever they started as.

> the synchronization that I want?

I am going to guess that you've fallen for the same thing I did at first: 
confusing "git pull" with "git fetch"; "git pull" can get you in 
trouble, "git fetch" probably won't.


Andy
-- 
Dr Andy Parkins, M Eng (hons), MIEE