I have two unrelated (not sharing any ancestor check in) Git repositories, one is a super repository which consists a number of smaller projects (Lets call it repository A). Another one is just a makeshift local Git repository for a smaller project (lets call it repository B). Graphically, it would look like this
A0-B0-C0-D0-E0-F0-G0-HEAD (repo A)
A0-B0-C0-D0-E0-F0-G0-HEAD (remote/master bare repo pulled & pushed from repo A)
A1-B1-C1-D1-E1-HEAD (repo B)
Ideally, I would really like to merge repo B into repo A with a single history timeline. So it would appear that I originally started project in repo A. Graphically, this would be the ideal end result
A0-A1-B1-B0-D1-C0-D0-E0-F0-G0-E1-H(from repo B)-HEAD (new repo A)
A0-A1-B1-B0-D1-C0-D0-E0-F0-G0-E1-H(from repo B)-HEAD (remote/master bare repo pulled & pushed from repo A)
I have been doing some reading with submodules and subtree (Pro Git is a pretty good book by the way), but both of them seem to cater solution towards maintaining two separate branch with sub module being able to pull changes from upstream and subtree being slightly less headache. Both solution require additional and specialized git commands to handle check ins and sync between master and sub tree/module branch. Both solution also result in multiple timelines (with subtree you even get 3 separate timelines when using --squash).
The closest solution from SO seems to talk about "graft", but is that really it? The goal is to have a single unified repository where I can pull/push check-ins, so that there are no more repo B, just repo A in the end.
See Question&Answers more detail:
os 与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…