by AJRF 3 days ago
I have never got a good answer to "can't you just make smaller PRs". This is convoluted tooling (needs its own CLI) for something you could achieve with just learning how git works.
IMO this tool is basically allowing you to do that, it just takes care of the bookkeeping to associate the series of smaller PRs with eachother which is possible today but requires a lot of clicking.
If there is a stack of size n and you make a modification at the first change, closest to the trunk, is there a single git command you can run to rebase the other n-1 branches and ensure they remote branches are updated?
Not a single one, but it can be done with 2.
Assuming you're currently on the most recent branch (furthest from the trunk), `git rebase -i --update-refs trunk` will rebase all the intermediate branches. If you need to make a change to the first branch nearest the trunk, either use `edit` in the interactive rebase, or make a fixup commit and enable autosquash for the rebase. The `--update-refs` flag makes sure that all the intermediate branches get updated during the rebase.
Then, to push them all, something like `git push origin 'refs/heads/yourname/*'` will push all branches prefixed with `yourname/`. It's a bit stupid that one can't just do `git push 'yourname/*'` though.
I agree that a `gh stack` command is not needed, but this feels to me like just a better UI feature for a good git workflow. It literally is about making multiple smaller PRs that build on top of each other.
The question is, why are you not just merging them into main as you go? It's a bit of a smell when you "need" to merge branches into branches. It shows a lack of safety and ease in deployments, which is the real problem to solve IMO.
Because sometimes there are changes that need to land as all or nothing.
Stacked PR’s… don’t do that though? They’re just PR’s. You can merge just the first one in the stack, and now it’s not “all or nothing”. Reading the docs, I don’t see a way to signal that the PR’s must all merge together.
Because the most natural way of saying “these changes need to land atomically” is called a branch, and landing it atomically is called a “merge”. But I guess GH’s UI sucks for reviewing large changes, so we’re stuck having to make each change independently mergeable and pass tests (likely disabling dead-code lints, etc) just to work around this limitation. Sigh.
At least when I actually do want changes to be mergeable in a stack, I now have a better UX for having folks review them.
Right, so 'just work the way the tool requires instead of making the tool work the way you want'. I would prefer the tool worked the way I want and the way I think of the changes instead.
I still don't have a single answer in this thread as to how even using this new tool even helps do things I can easily do with git?
> "Sometimes you just need to land changes all together".
Smell. What is that won't land? A UI change needed to go with a Database migration? That just tells me you don't flag your changes which you _should_ be doing. If you have code in your hot path that can break because of a DB or API change and you _don't_ flag that you have made your releases much more dangerous. Fix that before installing a new cli tool
If you can't do a database migration, you have put non optional new fields in somewhere, which doesn't accurately model your data domain - because it didn't exist before, it is tautologically optional.
> "How do I update a bunch of small PRs"
Smell. You should be merging small, flagged commits to main and rebase other branches. The workflow described is the same as feature branches - you start a branch, then branch from that, then branch from that - you shouldn't do that. That is not a git problem in the same way throwing every file on your desktop isn't an OS problem - you are just making a mess.