You must log in or register to comment.
I would dislike the first/referenced commit description as verbose as well. It describes a user or change drafting journey without ever saying concisely or separately what the commit actually does and why. If it at least had that summary up top in a first block or separated with
separator it’d be much better.I like the first part of the suggested alternative but I would never put a discovery journey into the commit message, or a “an hour of my life wasted”. I would put them in a MR comment - or separated block in the MR description with the intention of it not becoming part of the merge commit description.
The journey is not relevant to the code and changes. When you think of looking at it one year later, you can see the value of a description of the change, but I don’t see value in the discovery journey. The journey is more relevant in team-knowledge and workflow of how to work with the code base, and inter-personal team building.
Too much bloat of irrelevant information diminishes discoverability and conciseness of descriptive and useful information. It’s noise.
I just keep my commit messages one-liners and elaborate more on pull requests, where there’s enough context to really justify the change.
Putting the message in git puts the information closer to the code, since the pr isn’t in git itself but instead the git forge. You can for example search the text of git messages from the git cli, or come across the explanation when doing
git blame. I sometimes write verbose commit messages and then use them as the basis for the text in the pr, that way the reviewer can see it easily, but it’s also available to anyone who might come across it when doing git archeologyIt obviously depends on your exact git workflow, but my last team had things setup so that the code content of a MR was automatically squashed on merge, and the text if the MR itself was automatically set as the content of the new singular git commit.
This was largely the best of both worlds because your commits could have almost any text, and the description of what changed could be updated as needed when making the MR. But it ultimately ended up in the git history where it belonged.
Of course, I still had some trouble trying to get the team to describe their changes well in the MR at times - but that’s a different problem entirely.
Always squashing is a bit much for my taste, sometimes the individual commits have interesting information! Text from the MR in the merge commit is great though, maybe I should see if we can set that up with gitlab and propose that we start doing that at work.
I’m fine with squash merges for one commit. But otherwise, I consider structuring changes into commits structure too.
My team merges with merge commits which hold the MR description as a commit description, and MR title as commit title.
Individual commits are retained and can describe individual changes, while the MR and merge commit describe the whole changeset.
It’s a very interactive-rebase-heavy workflow (for commit cleanups/structuring when changes are added in review), but it works very well for us.
Normally when I merge a PR I put the long PR message (if there is one) in the merge commit (again if there is one), rather than shitty
Merge PR from patch1that people seem to use.You can actually change the behaviour on GitHub to be sane: https://blog.mergify.com/how-to-change-the-default-commit-message-on-github/amp/
If I’m not keeping the branch (usually PRs are not big enough to make preserving multiple commits useful) then I squash & merge which gives you the chance to edit the commit message and copy details from the PR message in.
I hate the default merge commits. I got quite frustrated when a FOSS project rejected (or didn’t come to a conclusion) my proposal for merge commits to also follow the commit formatting guidelines.
The cherry on top is merge commits describing which branch is being merged. But the branch disappears with that merge. I consider it worthless. The branch name is a name of the drafting process. There is no value to it when it lands.
shitty Merge PR from patch1 that people seem to use
One of the reasons I hate merge commits and just force linear history on the repos I control.
I think you could have just added a TLDR in the beginning of the commit message
deleted by creator
At work we always put an issue number (together with a short summary text) in the commit message. The the verbose explenation is done in the issue, where formatting, search and filtering is easier. However this creates a dependency to that issue system.
I would hate to have to open Jira tickets to be able to understand code or code changes in Git history.
I guess it depends on the quality of the summary. And how usable your issue system is.
I’m sure I would continue to feel uncomfortable though, about the code and history itself not being self-describing. Issue systems change, become unreachable, etc.
yeah, trying to convince our product owner to read commit messages wouldn’t go smoothly in my team. Some of them love their tickets…
the dependency to the issue system is unfortunate though, I do miss a “decentralized issue system” doing what git does for version control that would just seamlessly integrate with git and have a nice web GUI for less technical folks.
That’s called a mailing list
/s
god, I already struggle trying to find information in github issues, I can’t imagine using email for this
You don’t have to imagine it; you can browse the Linux Kernel mailing list!
That was rather insightful. And I definitely need a primer on writing good commit messages, and possibly on what a good version control commit is in the first place.
And TIL you can actually use some kind of markdown in git commit messages!
You can probably write your commit messages in most markup languages, whether it gets nicely rendered like it does here is a whole other topic
Me using stars
*for italic which renders italic in GitLab but as bold in TortoiseGit
