Update action should use `git pull --no-rebase` for integrating upstream changes

[ndonald2]

Some developers (including myself) prefer to use the "always rebase" configuration option for git pull, which avoids creating superfluous merge commits from un-pushed commits when pulling to a local branch from the same corresponding remote branch. This doesn't cause issues as long as you pull to the same corresponding branch, since the local un-pushed commits don't have remote counterparts yet and commit hashes won't be rewritten.

However, this does cause issues when pulling from a remote branch that is different from the local destination branch, since it's effectively rebasing local changes onto a totally different remote HEAD, which can cause commits to be rewritten.

I propose changing git update and any other command that attempts to pull from a different remote branch to use

git pull origin <upstream_branch> --no-rebase

or alternatively

git fetch origin <upstream_branch>
git merge origin/<upstream_branch>

since that's effectively what the pull should be doing.

[wireframe]

@ndonald2 great point. do you think it makes sense as a gitx default or is this more of a user level default?

git config --global pull.rebase true

https://coderwall.com/p/tnoiug/rebase-by-default-when-doing-git-pull

[ndonald2]

Hmm, I think there's a misunderstanding here. I personally do use the user-level git default for rebase pulling, as you described.

The issue is that gitx's current implementation doesn't take that into account - it uses whatever the user level default is for pulling, which breaks things if the default is rebase pulling. In the case of git update, the assumption as-implemented is that pulling from an upstream branch (e.g. pulling master into the current feature branch) will cause a merge, which does not hold true for the case when the default pull configuration uses rebasing.

I'm suggesting to change the gitx command invocations to forcibly opt-out of rebase pulling, using either method I described.

[ndonald2]

I might actually take a stab at a pull request for this. It would be a good ruby refresher :)

[f1sherman]

Fixed in #72