Re: [Gnu-arch-users] merge trouble

[Karel Gardas]

On Wed, 2 Mar 2005, Robert Anderson wrote:

> I've always gone out of my way to keep my branches related in simple
> ways so as to avoid merge troubles, but it looks like I've got things
> in a bind this time.
>
> I have two branches A and B.   B is a feature branch which is intended
> to fully merge into A.  (The situation is actually much more complex
> than this, but I HOPE this is enough to describe the problem.)
>
> Three times in the history of these branches,  I have done single
> patch one-off replay merges ("cherry picking") of a patch from B to A,
> when I'd made small fixes in the course of work on B that were needed
> right away in A (I know, there's a better way, but that's what I did.)
>
> Now I would like to merge to get B fully up to date with A, in
> preparation for the final merge of everything on B into A.
>
> But "star-merge A" in B's tree doesn't do at all what I want.  It ends
> up deleting most of my changes on branch B.  This is of course not
> what I want, and I admit to not fully understanding why this happens.
>
> So my question is:  what's the best way to proceed now?  Is there a
> way to effectively "back out" my cherry picking so that I can star
> merge the whole thing smoothly?  Or is there a better way?

I'm in the similar situation or very similar: I have two branches and done
several time in the history cherry-picking of several patches from A to B
and several times the same in reverse direction (with different patches of
course). Now, I'm just replaying what I can and omiting patches which are
results of merges (nice option is --skip-present). Anyway, the situation
is even more complex since both branches spread three archives in a
history and my tla (1.2 version) is completely blind w.r.t. star-merge and
crossing archive boundary. (this last sentense is quite unrelated issue
description, since star-merge is useless in my situation anyway...)

Cheers,
Karel
--
Karel Gardas                  address@hidden
ObjectSecurity Ltd.           http://www.objectsecurity.com