CODEWITHSUNDEEP
gitgit-rebasegit-merge-conflict
Stefan Witzel
Stefan Witzel

Reputation: 341

git rebase --rebase-merges produces merge conflicts

I have a git repository that has two branches with unrelated history like this

A'

A - B - [...] - C

where the [...]-part is complicated (non-linear and many commits). Both the commits A and A' are very simple, consisting of a single (nearly identical) file and commit B consists in removing that file and starting from scratch. Now I try to do a

git rebase --rebase-merges --onto A' B C

expecting that it will give me

A' - B' - [...] - C'

where X' is X with changed parent. I thought this would be fully automatic except possibly when creating B'. Instead git comes up with strange conflicts (allegedly removed files) even in non-merge commits that it asks me to resolve manually, but also asks me to re-resolve every merge-conflict that every occurred in [...].

In principle I would just like git to produce a commit B' identical to B except that it has parent A' (and a different SHA of course) and iterate. Is that not possible?

[In case someone is wondering, why I am trying to do this: Overleaf has a restricted git implementation where I can't just use an existing git repository to initialize a project but rather have to create an empty project (my A') there and then transplant my existing project on top of it.]

Upvotes: -1

Views: 66

Answers (2)

LeGEC
LeGEC

Reputation: 52081

Try running:

git rebase --rebase-merges --onto A' A C

In the following command: git rebase --rebase-merges --onto A' B C, the B commit would be excluded from the rebase -- much like B doesn't show up in the range git log B..C.

Perhaps the unexpected merge conflicts you witness come from the missing B' in the rewritten history ?

Upvotes: 3

eftshift0
eftshift0

Reputation: 30297

You could then, instead do:

git checkout A'
git cherry-pick B
# if there are conflicts there, it is because of the minor differences between the 2 files
# given that in B you remove that file, you could do this to wrap it up:
git checkout B -- .
git commit -m "Committing B on top of A'"

At this point B should be exactly like HEAD (check with git diff B) and then you should be able to pull the other stuff on top of B by using this script of mine: https://github.com/eantoranz/git-duplicate

./git-duplicate B C

It should give you a commit ID that has history just like what you want.

Upvotes: 2

Related Questions