On Thu, Sep 24, 2015 at 11:05 AM, Jianfeng Jia <[email protected]> wrote:
> If we can somehow enforce the rebase before merge, then I think it won’t > have the merge “tree” problem. > Actually the merge-tree problem was introduced in a situation where a part of a branch was cherry-picked to master. It would also happen if part of a branch were cherry-picked over to another branch (like you wanted to do with Taewoo's branch) and then both were eventually merged to master. In a normal scenario this wouldn't happen, with or without rebasing the feature branch before merge. As for whether "rebase before merge" should be enforced, I'm in favor of it personally. I'm even in favor of doing local interactive rebases with selective squashing and so forth to "clean up" local branch history before merging - although as always, this is destructively rewriting history and therefore should not be done if the branch was shared with anyone else. Ceej aka Chris Hillery
