My unorthodox, branchless git workflow April 6, 2020 on Drew DeVault's blog

I have been using git for a while, and I took the time to learn about it in great detail. Equipped with an understanding of its internals and a comfortable familiarity with tools like git rebase — and a personal, intrinsic desire to strive for minimal and lightweight solutions — I have organically developed a workflow which is, admittedly, somewhat unorthodox.

In short, I use git branches very rarely, preferring to work on my local master branch almost every time. When I want to work on multiple tasks in the same repository (i.e. often), I just… work on all of them on master. I waste no time creating a new branch, or switching to another branch to change contexts; I just start writing code and committing changes, all directly on master, intermixing different workstreams freely.1 This reduces my startup time to zero, both for starting new tasks and revisiting old work.

When I’m ready to present some or all of my changes to upstream, I grab git rebase and reorganize all of these into their respective features, bugfixes, and so on, forming a series of carefully organized, self-contained patchsets. When I receive feedback, I just start correcting the code right away, then fixup the old commits during the rebase. Often, I’ll bring the particular patchset I’m ready to present upstream to the front of my master branch at the same time, for convenient access with git send-email.

I generally set my local master branch to track the remote master branch,2 so I can update my branch with git pull --rebase.3 Because all of my work-in-progress features are on the master branch, this allows me to quickly address any merge conflicts with upstream for all of my ongoing work at once. Additionally, by keeping them all on the same branch, I can be assured that my patches are mutually applicable and that there won’t be any surprise conflicts in feature B after feature A is merged upstream.

If I’m working on my own projects (where I can push to upstream master), I’ll still be working on master. If I end up with a few commits queued up and I need to review some incoming patches, I’ll just apply them to master, rebase them behind my WIP work, and then use git push origin HEAD~5:refs/heads/master to send them upstream, or something to that effect.4 Bonus: this instantly rebases my WIP work on top of the new master branch.

This workflow saves me time in several ways:

I know that lightweight branches are one of git’s flagship features, but I don’t really use them. I know it’s weird, sue me.

Sometimes I do use branches, though, when I know that a workstream is going to be a lot of work — it involves lots of large-scale refactoring, or will take several weeks to complete. This isolates it from my normal workflow on small-to-medium patches, acknowledging that the large workstream is going to be more prone to conflicts. By addressing these separately, I don’t waste my time fixing up the error-prone branch all the time while I’m working on my smaller workstreams.

  1. I will occasionally use git add -p or even just git commit -p to quickly separate any changes in my working directory into separate commits for their respective workstreams, to make my life easier later on. This is usually the case when, for example, I have to fix problem A before I can address problem B, and additional issues with problem A are revealed by my work on problem B. I just fix them right away, git commit -p the changes separately, then file each commit into their respective patchsets later. ↩︎

  2. “What?” Okay, so in git, you have local branches and remote branches. The default behavior is reasonably sane, so I would forgive you for not noticing. Your local branches can track remote branches, so that when you git pull it automatically updates any local tracking branches. git pull is actually equivalent to doing git fetch and then git merge origin/master assuming that the current branch (your local master) is tracking origin/master. git pull --rebase is the same thing, except it uses git rebase instead of git merge to update your local branch. ↩︎

  3. In fact, I have pull.rebase = true in my git config, which makes --rebase the default behavior. ↩︎

  4. “What?” Okay, so git push is shorthand for git push origin master, if you have a tracking branch set up for your local master branch to origin/master. But this itself is also shorthand, for git push <remote> <local>:<remote>, where <local> is the local branch you want to push, and <remote> is the remote branch you want to update. But, remember that branches are just references to commits. In git, there are other ways to reference commits. HEAD~5, for example, gets the commit which is 5 commits earlier than HEAD, which is the commit you have checked out right now. So git push origin HEAD~5:refs/for/master updates the origin’s refs/for/master reference (i.e. the master branch) to the local commit at HEAD~5, pushing any commits that upstream master doesn’t also have in the process. ↩︎

Articles from blogs I read Generated by openring

Status update, May 2024

Hi! Sadly, I need to start this status update with bad news: SourceHut has decided to terminate my contract. At this time, I’m still in the process of figuring out what I’ll do next. I’ve marked some SourceHut-specific projects as unmaintained, such as…

via emersion May 21, 2024

Automatic case design for KiCad

I don't generally get along great with CAD software with the exception of KiCad. I guess the UX for designing things is just a lot simpler when you only have 2 dimensions to worry about. After enjoying making a PCB in KiCad the annoying for me is alwa…

via BrixIT Blog May 15, 2024

The floor is lawa!

And now for something completely different… When was the last time you were excited about a simple window with nothing but a single background color? Well, I currently am. Let me tell you about it… This window is notable, because it was created using the ”pu…

via blogfehler! May 8, 2024