Archiving Git branches as tags
etc.octavore.com100 points by octavore 3 days ago
100 points by octavore 3 days ago
On a related note, git supports incorporating custom commands via executables available in the `PATH` having the naming convention of:
git-<command name>
#!/bin/sh
# git-archive-branch
git_branch="${1:-$(git branch --show-current)}"
git co main &&
git tag archive/$git_branch $git_branch &&
git branch -D $git_branch
git archive-branch ...Seems like a sensible way to archive branches. It's not like a tag and branch are actually very different anyway, right? A tag is a pointer that always refers to the same commit while a branch is a pointer that follows commits forward. And for something that's archived, you don't need the pointer updating mechanism.
> A tag is a pointer that always refers to the same commit
It's not guaranteed not to change. The UI just makes it harder to update.
And anyone whose coworkers replace tags on a regular basis will be familiar with the following message from git when pulling changes:
would clobber existing tag
Really wish my coworkers would leave old tags as they were heh.
A hook should be able to prevent that
Hooks only keep honest people honest :) and an LLM will happily clobber a tag and skip hooks while the user just spams “accept”.
Can't solve culture problems with technology, but we all know that by now, right?
Luckily commonly used forges for collaboration have the ability to make tags immutable. Any repository where multiple people collaborate on a project should have that feature enabled by default. I'm still waiting for the day where tags are immutable by default with no option exposed to change it.
I'm sure that would cause problems for some, but transitive labels already exist in Git: branches.
I dont find the idea of a immutable "descriptive" tag or branch to be that useful (I also dont find the differentiation of tags and branches to be useful either) I've seen plenty of repositories where tags end up being pretty ambiguous compared to each other or where "release-20xx" does not actually point to the official 20xx release. Immutable references are more typically handled by builders and lockfiles to which Git already has a superior immutable reference system, the commit hash.
I 100% agree on the latter (the tag != release is more of a project management issue), and the same concept applies to containers and their digest hashes. The main issue at the end of the day is the human one: most people don't like looking at hashes, nor do they provide context of progression. I would say "give both" and make sure they match on the end user side of things, but tags are the most common way (open source) software releases are denoted.
As long as we can create and delete tags, they will never be immutable, right?
The purpose of the forge is to be able to prevent this. Protected tags are usually a feature which provides a way to mark tags as untouchable, so removal would require a minimum level of trust to the repository on the platform. Otherwise, attempts to push tag deletions or changes for tags matching the protected pattern would be rejected/ignored.
Of course, the repository owner has unlimited privilege here, hence the last part of my prior comment.
Tags are just a text file with a name and the sha of the tag object (with the commit and some metadata/signatures as contents), last I checked. It's deliberately simple and thus almost impossible to actually lock it down in concrete terms.
Packed refs are a little more complicated but all of the storage formats in git are trivial to manually edit or write a tool to handle, in extremis.
That's the purpose of the forge platform, to provide a way to prevent changes to these files from being accept into the source repository. For example:
https://docs.github.com/en/repositories/configuring-branches...
https://docs.gitlab.com/user/project/protected_tags/
https://forgejo.org/docs/latest/user/protection/#protected-t...
That's true for local hooks, but neither a dishonest person nor an LLM can bypass a pre-receive hook on the server (as long as they don't have admin access).
Correct.
One is refs/heads/<name> and the other is refs/tags/<name>
Branches are expected to change. When a commit is authored, HEAD (the current branch) updates to that commit.
Tags are expected not to change (though they can).
Other difference (actually, more like a consequence of what you said) is that Git keeps reflogs for branches but not for tags
> Important note: the strange magic requires the official git completion script.
I dislike the official git completion script because it’s so slow when working with large repos. On macOS - because of vagaries of the way its file system works - with large repos it’s effectively unusable. Hint to completion script writers: if your “smart” completion ever takes seconds, it’s worse than useless and far worse than “dumb” filename completion.
How often did you go back to the archived tagches that are older than say, 6 months ? Seems very niche, unless dunno, there are no version tags in the repo.
Believe it or not, some people need to use software more than six months old.
True. At work our flow is to tag commits that we want to mark as release candidates and delete feature branches after their PRs are merged/declined. We've never had a need to archive branches.
It seems very useful for archiving branches that never got merged.
Sometimes I work on a feature, and it doesn’t quite work out for some reason or another. The branch will probably never get merged, but it’s still useful for reference later when I want to see what didn’t work when taking a second attempt.
Currently, those abandoned branches have been polluting my branch list. In the past I have cloned the repo a second time just to “archive” them. Tags seem like a better idea.
I don’t think I’ve ever returned to a branch that I can easily rebase on top of the main branch. And if I really wanted to, I’d prefer to extract a patch so that I can copy the interesting lines.
Any branch older than 6 months is a strong candidate for deletion.
I sometimes leave merged branches around for quite a while, because I squash them when I merge to master and sometimes when tracking down a bug the ability to bisect very handy.
What made you decide to squash when merging instead of leaving the commits in the history so you can always bisect?
Not GP, but we do the same. Branches become the atomic unit of bisection in master, but the need is extremely rare. I think because we have good tests.
We also keep merged branches around. This has never happened, but if we needed to bisect at the merged-branch level, we could do that.
I know squash-merge isn't everyone's cup of tea, but I find it to be simpler and clearer for the 99+% case, and only slightly less convenient for the remainder.
Wonder if it's worth squashing in the branch, merging to main, then immediately reverting.
Now the work is visible in history, branch can be deleted, and anyone in the future can search the ticket number or whatever if your commit messages are useful.
Dunno if it's worth polluting history, just thinking out loud.
if you're deleting branches then why would you need to archive them? What would you even archive if you're deleting them? My deeper question is why are you deleting them in the first place though?
IMO a cleaner way to do this is with a headless remote, either on disk or “backed up” on a server. `git push —-all` won’t delete refs on the remote, so you don’t have to do any additional work to record or recover them.
`git push —all backup` will record all of your refs and tags
If you are archiving branches in your own rep, prefix with `ar/` so you can grep -v to conceal them.
See also `git notes` to record metadata an against a commit without changing the commit