Skip to main content

Considering the Philosophy of Squashing

Editor’s Note: I don’t like having to squash or rebase commits into a linear history. I am just operating in an environment where this is a dictated requirement. The rest of the post continues with the assumption that this preference is non-negotiable.


I have a sort of “git philosophy” discussion to pose. Recently, I had a massive headache doing an interactive rebase because I have a habit of periodically merging upstream master/dev/main/whatever branch commits into my working branch to keep the work of resolving merge conflicts to a minimum.
Unfortunately, when you have the situation of:

a----b----c----d----e----f
                \         \
        g----h----i----j----k----l

doing something like git rebase -i g will try and rewrite these pulls (a/b/c and d/e/f), even if you leave them as “picked” commits, and you only mark your commits (g/h/i/j/k/l) as “squash”. I really, really hate the idea of effectively changing/rewriting other developer’s commits; it can screw up stuff like git logs and git blames.

My workaround, of sorts, as been to:

  1. checkout a new branch from the upstream (git checkout -b merge-temp develop)

  2. do a “squash merge” to that branch from my working branch (git merge --squash feature/my-branch merge-temp)

  3. reset my working branch to the resulting merge commit (git checkout feature/my-branch && git reset --hard merge-temp)

  4. Delete the temp branch (git branch -D merge-temp)

This basically does the following:

a----b----c----d----e----f
                \         \
       g----h----i----j----k----l

becomes

a----b----c----d----e----f
                          \
                           g'

commits g through l get “rolled up” into a new commit, g', which can be merged. As a nice side-effect, a git merge --squash produces a very nice “squash summary” of each commit with its hash, author, and commit message in the squash commit message.


Now, on doing some more reading about git best practices, this LWN article came to my attention: https://lwn.net/Articles/328436/. In it, it talks about how Linus Torvalds handles the Linux kernel merges:

I want clean history, but that really means (a) clean and (b) history. People can (and probably should) rebase their private trees (their own work). That’s a cleanup. But never other peoples code. That’s a “destroy history” So the history part is fairly easy. There’s only one major rule, and one minor clarification:

  • You must never EVER destroy other peoples history. You must not rebase commits other people did. Basically, if it doesn’t have your sign-off on it, it’s off limits: you can’t rebase it, because it’s not yours.

    Notice that this really is about other peoples history, not about other peoples code. If they sent stuff to you as an emailed patch, and you applied it with “git am -s”, then it’s their code, but it’s your history.

It seems he feels that (a) people should squash their local histories before committing upstream, and (b) you should not periodically merge from upstream. That makes sense when you’re dealing with hundreds of commits from dozens of contributors, but what about us mere mortal devs, who work with maybe 7 developers on a project at a time?

I’d love to hear other, more experienced devs' thoughts.