Re: Non fast forward use case in beast

[Owen Taylor <otaylor redhat com>]

Hey Stefan,

Thanks for the detailed explanation. I'm going to answer it very briefly
- anybody is allowed to delete branches, so you can delete the win32
branch and repush it.

This will send all the commits in the branch to svn-commits-list again,
but the same would be true if you pushed a rebase as a non-fast-forward
commit.

- Owen

On Thu, 2010-01-21 at 11:26 +0100, Stefan Westerfeld wrote:
> Hi!
> 
> Recently I was warned by git that I was about to push non fast forward changes
> to the gnome repo. Now, this was the first time I was about to do this in a
> real repo, so I took the time to read up in various git howtos. Finally I
> decided to do git push -f, because I think in my case its the sensible thing to
> do. Which produced an error message and directed me to
> 
> http://live.gnome.org/Git/Help/NonFastForward
> 
> After reading the document, I am still convinced that I am trying to do the
> right thing here.
> 
> Anyway, here is my use case:
> 
> I've ported beast to windows, and my changes are in a branch called win32. Now,
> I am the only person who actually writes code which is committed into the
> branch.  From time to time, the branch needs to be rebased, so that it always
> gets the changes that happen in master (so conceptually win32 is the set of
> changes needed to build beast under windows). And this I was trying to do:
> 
> (in win32 branch)
> git-rebase master
> (check/ensure that everything still compiles and works with both, the
> changes from master and the win32 branch)
> git push -f
> 
> This is where I get an error message. Anyway, you may want to know what the
> long term perspective for the win32 branch is. The idea is to merge selected
> bits and pieces to master (after review), while rebasing the win32 branch
> occasionally. So in half a year or so, the branch may only contain some
> (~5 or so) hacky or hard-to-merge commits (as opposed to ~50 commits in the
> branch now).
> 
> So I'd like to push non fast forward changes, and I think its the right
> thing to do.
> 
>    Cu... Stefan