[Kde-scm-interest] Meeting minutes

[Oswald Buddenhagen]

On Fri, Nov 13, 2009 at 12:12:06AM +0100, Robert Wohlrab wrote:
> Oswald Buddenhagen wrote:
> > the imported history is completely rebased on top of the target
> > module. that way it is impossible to create an even remotely
> > consistent checkout of a past state where the two modules were still
> > separate.
>
> Wrong. If you do it with a subtree merge it is not a rebase, it is
> merged into the big repo with complete history. It looks then like the
> history of the one is only the history of the subfolder and onwards it
> is the history of both of them.
> 
i must admit a big hole in my git-fu here. what i've read now (the man
page, some how-to) doesn't really explain things in sufficient detail.
while i think i understand the state after the merge (there isn't
anything particular about it, after all), i don't see an obvious way how
to deal with the actual history. after all, there are two unrelated
commit histories which can be correlated by timestamp at best. that
means that i can checkout a past revision of the target repo and will
have no trace of the merged repo at that time. i might even be able to
check out a revision from the merged repo, but how would that fit into
the simultaneous checkout of the target repo?